Undergraduate Research Opportunities Program (UROP) Proceedings 2021-22

Abstract

UROP THE HONG KONG UNIVERSITY OF SCIENCE AND TECHNOLOGY

Proceedings 2021-22 Undergraduate Research Opportunities Program

Table of Contents

Message from the President i Sharing by Students ii-iii UROP Overview 2021-22 iv Mr. Armin & Mrs. Lillian Kitchell UG Research Awardees 2022 v-vi UROP Sponsorship Recipients 2021-22 vii Abstracts of UROP Projects – 2021-22* School of Science (CHEM, LIFS, MATH, OCES, PHYS) 2-68 School of Engineering (CBE, CIVL, CSE, ECE, IEDA, ISD, MAE) 70-163 School of Business and Management (ACCT, ECON, FINA, ISOM, MARK, MGMT) 166-197 School of Humanities and Social Science (HUMA, SOSC) 200-214 Interdisciplinary Programs Of f ice (EMIA, ENVR, PPOL) 216-228 Institutes and Centers (IEMS) 230

* Abstracts from each School are listed first by alphabetical order of the Department code, and then by Advisor’s surname.

Summary of UROP Courses

UROP 1000 Undergraduate Research Opportunities Program (0 credit with stipend option, offered in summer sessions only) UROP 1100 Undergraduate Research Opportunities Program Series 1 (1 credit, offered throughout the year) UROP 2100 Undergraduate Research Opportunities Program Series 2 (1 credit, offered throughout the year; prerequisite is pass in UROP 1100, with approval by project advisors) UROP 3100 Undergraduate Research Opportunities Program Series 3 1 credit, offered throughout the year; prerequisite is pass in UROP 2100, with approval by project advisors) UROP 4100 Undergraduate Research Opportunities Program Series 4 (1 credit, offered throughout the year; prerequisite is pass in UROP 3100, with approval by project advisors)

Message from the President

Since its launch in 2005, the Undergraduate Research Opportunities Program (UROP) has garnered a strong reputation within the community of the Hong Kong University of Science and Technology (HKUST). As one of HKUST’s signature programs, UROP has been successfully engaging our undergraduate students in research, providing early hands-on research experience alongside world-class researchers.

This volume of proceedings showcases the program’s accomplishments this past academic year and covers a wide range of UROP projects conducted by our undergraduate students under the guidance of their faculty supervisors. Read through these proceedings, and you will better understand UROP and learn about the numerous research ideas developed by our young students.

During the 2021/22 academic year, 143 faculty members served as project supervisors for more than 570 projects for the 602 students who participated in UROP. I would like to express my sincere gratitude to those faculty supervisors for generously taking on responsibilities beyond their regular teaching activities. By doing so, they are helping to build a vibrant research culture at HKUST in which our undergraduates can explore their research interests and thrive. I am proud that one-third of our UROP students have continued their research journey by pursuing graduate studies.

Building upon the foundation of excellence we have created in the 30 years since HKUST was established, the University will continue providing diverse and stimulating learning experiences to foster young, future researchers for the betterment of society and the world. As such, I am pleased that UROP will be further developed by being included in the University’s Common Core Framework in 2024. I also hope for your continued support and participation in the pilot of the new UROP course to be launched in the coming semesters. Please look out for the office’s announcement.

Sincerely, Professor Nancy IP President HKUST

i

Sharing by Students

CHANG Cheuk Yan Atta BEng in Bioengineering

UROP was the most valuable course throughout my HKUST journey. It allowed me to explore different fields and reinforce my career goal. Being an absolute UROP fan, I have taken three different UROP projects, each of which gave me the precious opportunity to learn and grow as a mature researcher. HKUST has a huge network of excellent supervisors and researchers. On top of learning laboratory techniques, they taught me the right attitude to fight challenges we encountered. Research is never about random trials and errors. I have learnt how to make hypotheses and systematically justify them. I was beyond grateful that UROP gave me the chance to explore what I am truly passionate about, as well as inspired me to become a researcher that can contribute to the betterment of society in the future.

CHEN Zehan BSc in Physics

To me, the 3-year UROP experience was priceless in showing me what exactly scientific research is. Although doing research seems out of reach for an undergraduate, do not be afraid. There is no need to know everything. You just need to learn a certain topic, be familiar with previous research work, and make your own improvement. In my UROP career, I encountered numerous difficulties, and felt anxious and tired from time to time. However, when I finally solved a problem, it was worth the effort. These are the valuable treasure and the fun of innovation and research that UROP brings to everyone who joins it.

Ii

CHENG Chun Hong BEng in Computer Science

UROP is one of the best opportunities provided by HKUST for students who wish to gain hand-on research experience and dive deeper into the academia. I fully enjoyed participating in this program. It allowed me to learn more about cutting-edge technology and how to write an academic paper to illustrate the exciting ideas started from 0 to 100. Not only research techniques, I also learnt how to collaborate with different colleagues within a research group to discuss and brainstorm ideas. I believe this program has equipped me well for my postgraduate study. For students who want to know more about academic research, please do not hesitate to participate in this program.

YAP Shan Qi BSc in Chemistry

Having no prior experience in research, I figured that participating in UROP would be a great way for me to start exploring my interests in research. Indeed, being able to really work on research projects has been a fruitful experience, as it has not only allowed me to gain insights into what research is about, but has also taught me skills and knowledge that I would not have learnt in the classroom. Besides, it has provided me a wonderful opportunity to work together with people who are passionate about research. Overall, I have benefitted tremendously from this research experience, and I am so glad to have chosen UROP to kickstart my journey in research.

Iii

UROP Overview 2021-22

The 2022 Mr. Armin and Mrs. Lillian Kitchell Undergraduate Research Award and the UROP Faculty Research Award

In the 2021-22 academic year, the Undergraduate Research Opportunities Program (UROP) has seen a substantial growth in participation compared to the previous year. Over 210 faculty supervisors from four Schools and from the Interdisciplinary Programs Office offered around 580 projects for our undergraduates. We also received more than 950 applications throughout the academic year, with 602 students were successfully enrolled to the program.

To enhance the continuous development of existing UROP projects, the UROP Support Grant Scheme continued providing financial support for faculty supervisors and their students. Through the Scheme, students also experienced the process of applying for research grants by submitting a joint application with their supervisors. In 2021-22, 38 out of 47 received applications were awarded with an aggregate funding amount of over HK$670,000, after the review by the UROP Office and by the UROP Advisory Board.

In the year 2022, 34 students with outstanding research performance were nominated by their UROP supervisors for the Mr. Armin and Mrs. Lillian Kitchell Undergraduate Research Awards. Among the nominees, 12 candidates were shortlisted and invited to give an oral presentation on their research findings to the UROP Advisory Board. The Board recommended 1 Champion, 2 First Runner-Ups and 3 Second Runner-Ups. The faculty supervisors of those 6 student awardees were also recognized by the UROP Faculty Research Award. The award presentation ceremony was held via Zoom on 29 April 2022.

Iv

Mr. Armin and Mrs. Lillian Kitchell

Undergraduate Research Award 2022

LIST OF AWARDEES

Champion & Best Poster Award

CHEN Zehan Major / Year: PHYS / 4 Supervised by: SHAO Qiming / ECE Project Title: Rational Design of Magnetic Devices Using Micromagnetic Simulations

First Runner-up

YAN Qiaolin Major / Year: CHEM / 4 Supervised by: SUN Jianwei / CHEM Project Title: Development of New Catalytic Organic Processes

First Runner-up

YAP Shan Qi Major / Year: CHEM / 4 Supervised by: SU Haibin / CHEM Project Title: Dynamic Expedition of Leading Mutations in SARS-CoV-2 Spike Glycoproteins

v

Second Runner-up

CHANG Cheuk Yan Atta Major / Year: BIEN / 4 Supervised by: CHAU Ying / CBE Project Title: Nucleic Acid Delivery by Self-Assembled Nanocarrier

Second Runner-up

CHENG Chun Hong Major / Year: COMP / 4 Supervised by: SO Hau Yue Richard / IEDA Project Title: Remote Vital Signs Monitoring

Second Runner-up

ZHOU Yangbo Major / Year: BCB / 3 Supervised by: PARK Hyo Keun / LIFS & PHYS Project Title: Real-time Imaging of Single Motor Proteins

vi

UROP Sponsorship Recipient 2021-22

The UROP sponsorship scheme is intended to provide UROP students with financial support to publish their papers in international journals, to present their posters or papers at academic conferences, or to participate in research-related summer schools or workshops during their undergraduate studies.

One student who has been awarded the UROP sponsorship in the 2021-22 academic year is listed as follows:

UROP PUBLICATION SPONSORSHIP

CHENG Chun Hong Major: COMP Supervised by: SO Hau Yue Richard / IEDA Publication: MDPI Sensors Issue: Sep 2021, Vol. 21(18) Paper: Deep Learning Methods for Remote Heart Rate Measurement: A Review and Future Research Agenda

vii

Abstracts of UROP Projects 2021-22

School of Science

School of Science Department of Chemistry

Department of Chemistry

Lead-free Perovskite Nanocrystals for Photo-induced Water Splitting

Supervisor: HALPERT Jonathan Eugene / CHEM Student: OUYANG Boyu / CHEM-IRE Course: UROP1100, Spring UROP2100, Summer

Colloidal photocatalytic hydrogen generation is a field that is gaining popularity in recent years. It allows conversion of solar energy into hydrogen gas to be stored as a fuel very easily. As a strong candidate in achieving large scale solar to hydrogen production, different photocatalysts have been synthesized. However, none of them achieved high enough efficiency or life spam long enough for commercial use. In this UROP project my work is to synthesize and study Cobalt(II) oxide (CoO) nanocrystal, a famous material for photo induced water splitting, and try my best to improve the system.

Construction and Application of Surface Enhanced Raman Spectrometer in Biomolecules Characterization

Supervisor: HUANG Jinqing / CHEM Student: CHAU Cheuk Yui Sherman / BICH Course: UROP1100, Fall

With its highly distinct, non-destructive characterization of chemical samples, Raman spectroscopy is utilized in a plethora of fields, including pharmaceuticals, cosmetics, and geology. Its rapid identification of samples could therefore prove useful in differentiating between cancerous and non-cancerous breast tissue from the same patient. This report presents the application and viability of surface enhanced Raman spectroscopy in the characterization of biomolecules. The results provide evidence that the biochemical composition of tumor, papilloma, and fibroadenoma are notably different but contain a similar spectra pattern. To analyse the spatial configuration of the tissue sample, Raman mapping is also utilized to visually identify the discrepancy between cancerous and noncancerous breast tissue and its influence on Raman scattering.

Construction and Application of Surface Enhanced Raman Spectrometer in Biomolecules Characterization

Supervisor: HUANG Jinqing / CHEM Student: ZHOU Daoyu / SSCI Course: UROP1100, Summer

In this research, we utilized Raman imaging combined with unsupervised machine learning (UMAP-DBSCAN) as an assistance to achieve the goal of determining the margin of the breast ductal carcinoma sample. So far, we have finished the Raman peak assignment and performed analysis based on the clusters obtained from DBSCAN, Raman imaging, and mean spectra of each cluster thereby some preliminary conclusions have been drawn. The clusters generated from machine learning show a well transitional pattern from collagens to malignant cells in many cases. We found malignant cells have higher lipids and protein contents and altered configurations in specific bands of Raman spectra compared with cells adjacent to ECM in the same region. Several possible reasons for these have been interpreted.

2

School of Science Department of Chemistry

Prediction of Inorganic Solar Cell Materials with Double Perovskite Structure Using Machine Learning Approach

Supervisor: SU Haibin / CHEM Student: CHEUNG Ka Key / CHEM Course: UROP1100, Fall

Solving the covalency and ionicity partition in solid states has many useful applications. For example, iconicity affects grain boundary which Jahn-Teller distortion can happen and directly determines the strength of the solid. In this work, we investigate CsPbI3, solid state material with cubic perovskite structure. Vienna Ab initio Simulation Package (VASP) is used for the simulation which gives useful parameters such as potential and electron density in different locations of the lattice. We can break down the energy into ionic and polarization energy partition to understand the ionicity.

Deep Learning in Synthesis Planning

Supervisor: SU Haibin / CHEM Student: CHUI Sin Yu / CHEM Course: UROP3100, Fall UROP4100, Spring

Nickel-catalyzed reactions raised more attention in the development of organic synthesis. The benefits of using nickel compound as the homogeneous transition metal catalyst for cross-coupling reactions have been recognized by different researchers. Cross-coupling reactions involving C-O bond activation, C-N bond activation and C-H bond activation have also been developed to form organic compounds. This report will be focused on using computational methods to collect research papers on homo-nickel-catalyzed crosscoupling reactions. Further analysis of photochemistry will also be mentioned.

Deep Learning in Synthesis Planning

Supervisor: SU Haibin / CHEM Student: DOO Kwan Lam / CHEM Course: UROP1000, Summer

Synthesis planning is essential for many reactions. Using this technique, it is possible to discover many new reactions to make different molecules in a faster and cheaper way. It would have many applications in drug discovery, medicinal chemistry, etc. This summer I learned about the synthesis planning of homogeneous nickel catalysis. This report is an overview of my data extraction progress in 2022 June, July, and August. The task includes data extraction of homogeneous nickel catalysis and some copper catalysis reactions. I will analyze the data I collected from the three aspects of leaving group, reactant group, and ligands with figures.

3

School of Science Department of Chemistry

Deep Learning in Synthesis Planning

Supervisor: SU Haibin / CHEM Student: WANG Yizhou / CHEM Course: UROP2100, Fall

Most of organic reactions require catalysts in order to occur and end efficiently, in which metal catalysts attract a lot of interest of researchers to study. Unlike heavy metals, which are rare and toxic, some light metals, such as nickel, can also perform wonderfully in catalyzing various kinds of organic reactions, like reductive coupling, C-O activation, etc. Besides, nickel is also capable of catalyzing the conversion of carbon dioxide into acetic acid in order to absorb the greenhouse gas. This report is an overview of reactions and conditions using nickel catalysts with bidentate nitrogen ligands, and will analyze the conditions including ligand types, base and so on.

Virtual Reality in Chemistry Supervisor: SU Haibin / CHEM Student: KONG Yui Hin / CHEM Course: UROP1100, Fall

Molecular dynamics (MD) refers to using computer simulations to study and predict physical movement of atoms and molecules over time. MD spans across multiple fields of science as a easy and efficient tool to predict or verify experiments. MD as a field of computational chemistry is becoming more popular as computing power becomes more cheaper and easily available. Here, a brief review of history and principles of molecular dynamics is given. Then, multiple MD scenarios are simulated on a consumer level computer to compare efficiency. Futures and limitations of MD are discussed.

The Impact of Spike Mutations on SARS-CoV-2 Neutralization

Supervisor: SU Haibin / CHEM Student: ANGELA Donna / SENG Course: UROP1000, Summer

The COVID-19 pandemic has been a global challenge people are facing. One of the main issues is the development of new variants, making it more dangerous for us. Despite many efforts given by various parties, the team and I involve in finding a novel way to track mutations of the spike protein, mainly RBD and NTD discussed here. One of the methods in fulfilling our goals is through variant decomposition, which gives access to specific mutation information of the unique sequences we have and discovers the significant mutation trend related to relevant amino acid types for variants of concern or interest targeted.

4

School of Science Department of Chemistry

The Impact of Spike Mutations on SARS-CoV-2 Neutralization

Supervisor: SU Haibin / CHEM Student: CHANG Ka Pui / CHEM Course: UROP1100, Summer

The COVID-19-causing virus SARS-CoV-2 have mutated frequently since its debut in December 2019, mainly in its spike glycoprotein. By having the combination of different types of amino acid mutations in the spike, for example, amino acids substitution on a site (SAP) which may have altering chemical properties, deletion and glycosylation, commonly N-linked glycosylation, the emerged variants have quite distinct viral properties such as a change in virulence or transmissibility. Here, sequence tracking is used to detect a pattern of mutations using representative sequences found before the emergence of VoC/VoI, to account for their viral properties, effects to the pandemic and to have a more accurate prediction for the next VoC/VoI.

The Impact of Spike Mutations on SARS-CoV-2 Neutralization

Supervisor: SU Haibin / CHEM Student: JUSMIN Frederick Gabriel / CHEM Course: UROP1000, Summer

The rapid mutation of SARS-CoV-2 leads to the emergence of new variants and numerous key mutations which are conserved in the viral genome. The positive selection of these mutations produce evolutionary adaptations allowing the virus to be more infectious and more virulent. As a way to track new mutations appearing in future sequences, a moving reference chosen from past sequences with highest genetic similarity is used instead of the original strain. Through observation of monthly plots, leading counts in mutated sites reveal some insights to key mutations which later appear in SARS-CoV-2 variants while counts of reference sequences shows the previous key mutations. Additionally, further analysis of these mutated sites will definitely be a contributing factor to determine the evolutionary trajectory of SARS-CoV-2.

The Impact of Spike Mutations on SARS-CoV-2 Neutralization

Supervisor: SU Haibin / CHEM Student: LAM Chak Fai / CHEM Course: UROP1100, Spring UROP2100, Summer

COVID-19 pandemic has been affecting the world for almost 3 years, continuous records of mutations of the acute respiratory syndrome coronavirus 2 (SARS-CoV-2) were made. Spikes proteins in the new variants are mutated with improvements in both transmissibility and immune escape, leaving the lives of billions in danger. Investigations and predictions of the spike glycoproteins of SARS-CoV-2 mutation, which is the key to the entrance into target cells, are done worldwide. Using the data from January 2022 to June 2022 in the GISAID hCoV-19 database, this report aims to compare different approaches that assist the tracking of emerging variants. Data of dynamic expedition of leading mutations (deLemus) and escape fraction are the two methods used to compare with the results of this report for further analysis.

5

School of Science Department of Chemistry

The Impact of Spike Mutations on SARS-CoV-2 Neutralization Supervisor: SU Haibin / CHEM Student: LEUNG Cheuk Fung Alvin / SSCI Course: UROP1100, Summer

The spike (S) glycoprotein of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is crucial for cell entry by binding to human angiotensin-converting enzyme 2 (hACE2). Despite being distal from the S-hACE2 interface, the two subunit 1 (S1) subdomains (SD1 and SD2) and subunit 2 (S2) play important roles in viral functions via allosteric modulation of hACE2-binding and mediation of membrane fusion respectively. Hence, mutations in these regions confer significant impacts to SARS-CoV-2 functions. Here, the time-resolved dynamic expedition of leading mutations (deLemus) method is applied to screen for sites of interest (SOIs) in aforementioned domains based on their weighted L-index scores derived from single amino acid polymorphism (SAP) analyses. These findings would allow the potential projection of S evolution trajectory.

The Impact of Spike Mutations on SARS-CoV-2 Neutralization

Supervisor: SU Haibin / CHEM Student: TSANG Kwok Kiu / CHEM Course: UROP1100, Spring

Amidst the CoVID-19 pandemic, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has undergone rapid, ongoing genomic evolution, forming numerous variants that have developed increased transmissibility and antibody evasion. The spike glycoprotein of the virus is crucial for antibody development and is also integral in the receptor recognition and cell membrane fusion process, hence monitoring and prediction of mutations for this protein is of importance. Here, we propose a method of identifying new active sites of mutation for the spike glycoprotein by employing a direct counting method, used in conjunction with statistic coupling analysis (SCA), for continuous monitoring of its mutation patterns. The identified new active sites of interest (SOIs) show promise in predicting the emergence of future variants of concern.

Text Mining of Synthesis Methods of Metal Organic Framework

Supervisor: SU Haibin / CHEM Student: SIU Chun Hey / SSCI Course: UROP1000, Summer

Metal Organic Frameworks (MOFs) has gained rising interests in recent studies for their huge internal surface area and astonishingly high porosity which, coupled with the extremely flexible nature of its organic and inorganic constituents, revealed countless possibilities for a plethora of novel applications. However, there is still a lack of understanding on the specifications of the synthesis of MOFs. This UROP project works on overcoming this limitation by extracting and evaluating synthesis methods and conditions from thousands of studies. Via an automation software that allows customization of document digitization workflow, Text Mining and Graph Mining are performed extensively on selected studies to gather a wide range of data for MOFs.

6

School of Science Department of Chemistry

Data Analytics of Homogeneous Transition Metal Catalyzed Reactions

Supervisor: SU Haibin / CHEM Student: WANG Yizhou / CHEM Course: UROP3100, Summer

Coupling reaction is a significant type of reactions in organic synthesis. In the past decades, researchers have found many transition metals that are able to catalyze various kinds of coupling reactions. As a cheap transition metal, nickel is thought to be an excellent metal to effectively catalyze coupling reactions, and many groups have made a lot of experiments and data about nickel catalyzed C-O activation reactions. This report will try to review different C-O activation reactions catalyzed by nickel and give some characteristics or rules about this reaction field.

Development of New Catalytic Organic Processes

Supervisor: SUN Jianwei / CHEM Student: HO Hang Chi / CHEM Course: UROP1100, Spring

Synthetic organic chemistry is commonly used to synthesize important molecules in our daily lives such as drug molecules and pharmaceuticals. However, many of these molecules contain a chiral carbon atom and have two non-superimposable mirror images, or enantiomers, which can have vastly different biological activity. Therefore, reactions that can preferentially form one enantiomer over the other has seen rapid development in the past century precisely because of its applications in the pharmaceutical industry. Such asymmetric reactions can be catalyzed by either enzymes, transition metal complexes, or organocatalysts.

Development of New Catalytic Organic Processes

Supervisor: SUN Jianwei / CHEM Student: LI Yuxuan / CHEM Course: UROP1100, Fall

Based on preceding results, further optimization of catalytic enantioselective 1,6-addition of 5-methide-5Hindoles, which are generated in situ from dehydration of corresponding indolyl alcohols, has been attempted. Starting from modification on structures of the substrates, different chiral phosphoric acids were used to catalyze the reaction. With the result of catalysts screening in hand, evaluation of common solvents, concentration of reactants, and temperature has been performed. As a continuation of the previous project, this work successfully enhanced the enantioselectivity of the reaction (ee value up to 78%). Although the enantioselectivity of such reaction mode requires further enhancement, the increase in ee value could still prove that optimization on the reaction is effective.

7

School of Science Department of Chemistry

Development of New Catalytic Organic Processes Supervisor: SUN Jianwei / CHEM Student: LO Wai Yam / CHEM-IRE Course: UROP1000, Summer

Catalysis plays a crucial role in the synthesis of biologically active molecules. Many drug molecules are synthesized via organic synthesis with the aid of catalysts where natural occurrences are much lower, and resources are scarce. While metal catalyzed reactions provide a powerful tool in drug synthesis, the recent development of organocatalysts provides new advantages over metal catalysts, including increased robustness, recyclability, decreased risk of metal contamination, and eliminated the need for consuming precious noble metals, while maintaining good selectivity and yield.

Making Valuable Organic Molecules with Green Chemistry

Supervisor: TONG Rongbiao / CHEM Student: HE Chenxi / CHEM Course: UROP1100, Fall

Alcohol oxidation is an important reaction in organic chemistry and many chemists have spent great efforts to develop various conditions for multiple complicated substrates. However, most popular methods used at current have drawbacks in environment-friendliness, safety and operation difficulty. Therefore, we aim at developing a new method with Fenton Chemistry to achieve low-cost reagents, non-hazard waste, broad scope and high reaction yield. In this report, background information of both the present alcohol oxidation methods and the Fenton reaction will be provided first, including some important references. Then the motivation and whole research plan of timeline will be briefly discussed. After that, the research progress till 25th November will be discussed in detail for three stages, including the mechanism prediction and examination, the reaction condition modification, and the substrate scope extension. Last but not least, the reflection and limitation, following by the future direction of this project will be presented at the end. The references for this reported will be listed in the last page.

Design of Controllable Enzymes for Therapeutic Applications Supervisor: VONG Kenward King Ho / CHEM Student: LAU Chi Hin / CHEM Course: UROP1100, Summer

Conventional inflammation-suppressing drugs mainly target the synthesis of PGH2 from arachidonic acid, which is catalyzed by cyclooxygenase-2 (COX-2). This project aims to explore the possibility of utilizing the rigid binding between MUC1-N and MUC1-C to regulate the activity of 15-hydroxyprostaglandin dehydrogenase (15-PGDH), the key enzyme controlling PGH2 catabolism. With such modification, the enzyme could be injected into sites of inflammation, leading to the controlled breakdown of PGH2. For this project, expression and purification of the modified 15-PGDH was performed.

8

School of Science Department of Chemistry

Design of Controllable Enzymes for Therapeutic Applications Supervisor: VONG Kenward King Ho / CHEM Student: WAN Kai Yui / IRE Course: UROP1000, Summer

Stroke is a serious and fatal disease, but antithrombotic agents can be applied to interrupt plaque formation to significantly reduce mortality rates. The protease known as nattokinase is also a viable antithrombotic agent, but may cause brain bleeding from excessive protease activity. In this UROP project, a modified nattokinase will be produced using recombinant DNA technology. Following expression in Escherichia coli, protein purification was performed using techniques like Fast Protein Liquid Chromatograph (FPLC), centrifugal filtration, and High-Performance Liquid Chromatography (HPLC). The goal of this project was to optimize purification conditions for these therapeutic proteins.

9

School of Science Division of Life Science

Division of Life Science Mechanisms of Golgi Membrane Protein Retention Supervisor: BANFIELD David Karl / LIFS Student: LI Caifan / SSCI Course: UROP1100, Summer

Vps74p provides a protein retention mechanism in Golgi to maintain the steady-state distribution of glycosyltransferases (GTs) and has always been an important topic in cell biology study. There have been many studies using fusion expression of fluorescent proteins assay showed that Vps74p bound both the binding motifs on the cytoplasmic tail of GTs and the COPI coat then achieved the protein retention in the Golgi. However, this method has limitations in the quantitative analysis of retained protein content. Thereby, in this report, we imitated a mammalian cell quantitative Golgi retention assay to demonstrate its applicability in yeast. However, our result showed when VPS74 was deleted, a reconstructed reporter protein with a Vps74p binding tail, which localized on the plasma membrane in wide-type cells (WT), could not go back to the plasma membrane as we expected but further accumulated in vacuoles. To improve the experiment, we excluded the interference of a mutation on the SI TMD, but the result was still not good. Finally, we offered several possible causes for this problem and a way to improve it.

Studying Novel Cell Cycle Signalling Pathways in Yeast Cells Supervisor: BANFIELD David Karl / LIFS Student: SINCE Alec Victor / BIBU Course: UROP2100, Spring

This literature review aims to investigate the role of TORC1 in regard to the various checkpoints in the cell cycle. This paper suggests that TORC1 is crucial for cell cycle progression due to its effect on the subcellular localization of a polo-kinase called Cdc5. Our lab used dcr2-6 ted1Δ yeast cells and made contradictory findings where we determined that TORC1 is required for spindle assembly checkpoint (SAC) activation at non-permissive temperatures in those strains. We came to this conclusion as the inhibition of TORC1 led to the inactivation of SAC. Further studies are required to determine whether Cdc5 also has an effect on dcr2- 6 ted1Δ yeast cells.

Studying Novel Cell Cycle Signalling Pathways in Yeast Cells Supervisor: BANFIELD David Karl / LIFS Student: YUK Tin Cheung / BTGBM Course: UROP1100, Fall

GPI-Anchors contain mannose molecules, of which Man2 is attached to EtNP (Glycosylphosphatidylinositol) normally by the action of enzyme Gpi7p. After the attachment of EtNP to the GPI-Anchor protein, the Ted1p/Dcr2p enzymes would do the opposite to detach the EtNP from these GPI-Anchor proteins. Previously, it was found out that on the mutant strain of yeast (PGAL1GPI7ted1Δdcr2Δ), which lost the ability to produce Ted1p/Dcr2p enzyme, a lethal effect was demonstrated. This somehow revealed the toxicity of EtNP on the survival of cells.

10

School of Science Division of Life Science

Molecular Regulation of Muscle Stem Cell Quiescence by Non-coding RNAs Supervisor: CHEUNG Tom / LIFS Student: CHEUNG Lok To Curtis / BCB Course: UROP1100, Summer

Muscle stem cells are often found in quiescence, a cellular state which exists in a dormant, reversible G0 cell cycle state outside of the cell cycle. Quiescent muscle stem cells could be activated and enter the cell cycle for proliferation or activation, fulfilling the tissue homeostasis and repair role of stem cells. It is important to understand the various governing mechanism by which quiescent stem cells rely on to switch between quiescent and activated states in order to further the research on the application of adult stem cell rejuvenation. In this report, we will discuss how autophagy, one of the proposed mechanisms could be correlated to the muscle stem cells' cell fate and their autophagy flux, using FACS isolated mice adult skeletal stem cells.

Molecular Regulation of Muscle Stem Cell Quiescence by Non-coding RNAs Supervisor: CHEUNG Tom / LIFS Student: CHIU Yui Hei / BIOT Course: UROP1100, Fall

M2 is a potential anti-cancer drug candidate studied in Dr Liang Chun’s laboratory. The drug candidate shows its exceptional potential and specificity in inhibiting the growth of cancer cells, particularly Hela cells. However, the long-term use of the drug exhibits a critical issue– Drug resistance. Previously, 2-8, 4-5 and 5- 6 clones from the R10 population (Highest drug resistance HeLa population available in our laboratory) were studied in depth through various assays, confirming their morphology. This project devoted itself into studying another two clones; 2-9, 4-4 in comparison with the HeLa population by conducting various assays. We found that 2-9 and 4-4 exhibited similar trends as the previous R10 clones with slower growth, higher survivability after drug treatment and smaller in size. Unlike the previous clones, 2-9 and 4-4 exhibited different behaviours implying different degree of resistance to the M2 drug.

Molecular Regulation of Skeletal Muscle Stem Cell Quiescence and Activation Supervisor: CHEUNG Tom / LIFS Student: TSE Chun Mong / IRE Course: UROP1100, Summer

Pax7 is a marker of satellite cells (SC)1, so in order to aid the study of SC, transgenic Tg:Pax7-nGFP mice was developed. Expression of nuclear localised EGFP (nGFP) transgene corresponds with that of the endogenous Pax7 gene2, thus marks SC with green fluorescent protein (GFP). However, GFP signals may not fully represent Pax7 expression6. Here, we would like to see to what extent does GFP marks Pax7, taking this to wrap up things done in this summer.

11

School of Science Division of Life Science 12 Molecular Regulation of Skeletal Muscle Stem Cell Quiescence and Activation Supervisor: CHEUNG Tom / LIFS Student: WONG Tsz Yan / BIBU Course: UROP1100, Summer

Quiescence and activation of stem cells have been previously proven to be correlated to autophagy, which is the process of self-degradation of proteins. Utilizing muscle stem cells, or satellite cells isolated from adult skeletal muscle from hindlimb of mice with fluorescence-activated cell sorter (FACS), the cell fate is examined from inhibition and promotion of autophagy. The autophagy granules, namely autophagosomes and autolysosomes are also evaluated with confocal microscope. As autophagy is closely related to ageing, the study of autophagy of satellite cell can provide insights for therapeutic development.

Construction of a Signal Transduction Pathway Reporter Indicator for Monitoring Signaling Strength Supervisor: CHOW King Lau / LIFS Student: YUEN Yan Yi Macy / BIOT Course: UROP3100, Fall

This project will manipulate the genetic model system, Caenorhabditis elegans, to recruit two of the gene components regulated by the transforming growth factor-β related ligand, DBL-1, as the indirect indicators of the dbl-1 signaling strength. sma-6, one of the target genes for the project, encodes the type I receptor for DBL-1 ligands and is positively regulated by the dbl-1 signaling. The second gene, gcy-28, is a receptor-type guanylate cyclase and acts as a repression target in the BMP pathway. Thus, the gene expression of sma-6 and gcy-28 can reflect the intensity of the BMP transduction signal. This project will develop the sma- 6 and gcy-28 transcription reporters responding to the dbl-1 signaling and amplify the readout of signal strength through the manipulation of these two reporters in a variety of dbl-1 mutant strains.

Monitoring the Norm of Human Morphology Supervisor: CHOW King Lau / LIFS Student: LUI Yuen Yee / SSCI Course: UROP1000, Summer

The purpose of this paper is to test whether the body shape of modern human beings can also be described with parameter defined by the “Vitruvian man”. Quantitative and qualitative data were collected through the measurement of online photos. The data collected indicate that most of the statements described on the Vitruvian man are not correlated to the experimental observations. Distinct differences were found when comparing the results of two races and two sex groups. These distinct differences further act as several defining parameters to distinguishing two different races and two sex groups. This paper and experimental results can be used in investigating the biological correlation of distinct body proportion difference.

12

School of Science Division of Life Science

Monitoring the Norm of Human Morphology Supervisor: CHOW King Lau / LIFS Student: RUPANI Rakshita / SSCI Course: UROP1000, Summer

This paper re-evaluates the facial proportions of the Vitruvian Man - a drawing constructed by Leonardo da Vinci in about 1490 that defines human body proportions - and determines how accurate it is to African, Caucasian, and Hispanic ethnicities today. The paper also deduces the numerical ratios that define certain facial features that allow differentiation between ethnicities. Measurements of facial features - in samples collected from the Internet - were done to calculate ratios. The ratios obtained from samples of different ethnicities (same gender) were compared using t-tests. Only some ratios allowed differentiation between ethnicities. Data from females gave more consistent measurements than men. Africans were less related to Caucasians and Hispanics, while Caucasians and Hispanics has a stronger correlation.

Genetic Identification of Negative Regulator of Bone Morphogenetic Protein Signaling Pathway Negative Regulator Supervisor: CHOW King Lau / LIFS Student: CHOI You Jin / BIOT Course: UROP1100, Summer

Body size is one of the most important phenotypic features in animal species. Although we understand that signaling pathways control body size, the molecular mechanism and how the downstream effectors are regulated are poorly understood. In Caenorhabditis elegans, a conserved Bone Morphogenetic Protein (BMP) signaling pathway, the Sma/Mab pathway, is the predominant regulator of body size. LON-1, a member of the cysteine-rich secretory protein (CRISP) family, acts as the negative regulator to suppress the small body size phenotype when mutations in lon-1 gene occur. To understand why lon-1 worms are long, we aim to understand whether the Lon phenotype of lon-1 mutant can be modulated by different collagen gene functions to control the body size of C. elegans.

Analysis of Surface Delivery of Epidermal Growth Factor Receptors Supervisor: GUO Yusong / LIFS Student: CHENG Guo / SSCI Course: UROP1000, Summer

In intracellular trafficking, Trans-Golgi Network (TGN) is the transportation hub that sorts proteins to various downstream destinations including endosomes, plasma membrane, etc. The surface delivery of Vangl2, a conserved signaling receptor regulating Planar Cell Polarity (PCP), has been one important research direction in understanding the elaborate sorting mechanisms. It was previously demonstrated that both AP1 and Arfrp1 are involved in the transportation of Vangl2 from TGN to the plasma membrane. (Guo, Zanetti & Schekman, 2013) Here we purified AP1 core to study the interaction between AP1 core and Vangl2, as well as the function of Arfrp1 in regulating the TGN export of Vangl2.

13

School of Science Division of Life Science

Analysis of Surface Delivery of Epidermal Growth Factor Receptors Supervisor: GUO Yusong / LIFS Student: HUNG Shing Hei / IRE Course: UROP1000, Summer

The analysis of EGFR is highly related to understanding the development of cancer cells. The L858R mutation on EGFR is one of the most common mutations identified in lung cancer. T790M mutation on EGFR causes drug resistance of EGFR(L858R). This project is trying to introduce the further mutation T790M on EGFR(L858R) for further biological analysis.

Analysis of Surface Delivery of Epidermal Growth Factor Receptors Supervisor: GUO Yusong / LIFS Student: LI Chun Wa / BCB Course: UROP1100, Fall UROP2100, Spring

Mitochondria has been extensively studied since its discovery of the organelle in 1857 by physiologist Albert von Kolliker, unravelling many features of this bacterial descendent, from its fission and fusion machinery, to its interactions with other organelles via signaling or membrane contact sites. Mitochondrial-derived vesicles (MDVs) is an emerging field that is comparatively less studied yet exciting in terms of understanding mitochondrial regulations and their physiological relevance. MDVs are responsible for diverse cellular processes, in which its role in mitochondria quality control (MQC), peroxisome biogenesis and mitochondria antigen presentation (MitAP). The molecular machineries controlling cargo selection, budding and trafficking of these processes will be discussed in this review.

Analysis of Surface Delivery of Epidermal Growth Factor Receptors Supervisor: GUO Yusong / LIFS Student: PARK Jihye / BIOT Course: UROP1100, Spring

Sonic Hedgehog (Shh), one of the three Hedgehog (Hh) family members, is an important signalling molecule involved in many key developmental processes, such as embryonic patterning, cell differentiation, and organ development. The Shh protein undergoes two major changes in the endoplasmic reticulum (ER). It is first auto-cleaved into two parts, a C-terminal fragment (ShhC) and an N-terminal fragment (ShhN). ShhC is degraded, while ShhN is exported from the ER after lipid modification and delivered to the cell membrane to initiate signal transduction. Although the signal transduction pathway is well-understood, the complete mechanism behind Shh protein secretion is still undiscovered. Therefore, this paper examines the transport of the N-terminal fragment of Shh without the lipid modification (ShhN) in vitro to unravel the molecular mechanisms that direct the trafficking and secretion of Shh.

14

School of Science Division of Life Science

Molecular Mechanisms Regulating Biogenesis of Peroxisomes Supervisor: GUO Yusong / LIFS Student: LI Chun Wa / BCB Course: UROP3100, Summer

Peroxisomes are vital organelles for most eukaryotes, while they undergo growth and division under homeostatic conditions, they can be generated de novo when depleted. Although the biogenesis of organelle occurs in the ER of yeasts and plants, peroxisomes in mammalian cells have an interesting dual origin from the ER and the mitochondria, in which peroxins Pex3 and Pex16 are the key players in the formation process. To unravel the underlying mechanisms regulating this event, we aim to utilize retention using selective hooks (RUSH) assay to study the trafficking machineries of these peroxins from these 2 donor organelles by synchronizing the vesicle release of these proteins. In UROP 3100, immunofluorescence was performed to determine the localization of Str-Ii_Pex16-SBPEGFP and Str-HA-MAO_VSVG-SBP-EGFP, the ER and mitochondria hook for Pex16 and Pex3, respectively. Our study shows that while VSVG-SBP-EGFP successfully retained in the mitochondria, Str-Ii failed to lock Pex16-SBP-EGFP in ER but instead allowed it to be transported to the ER.

Molecular Mechanisms Regulating Biogenesis of Peroxisomes Supervisor: GUO Yusong / LIFS Student: YAU Yin Lam / SSCI Course: UROP1100, Summer

In the eukaryotic secretory system, endoplasmic reticulum (ER) and Golgi apparatus play important roles. During the secretion process, coat protein complex II (COPII) captures cargo proteins at ER and form vesicles in a bid to transfer the proteins to Golgi apparatus. Previous studies have shown that PRRC1, a cytosolic factor, may be involved in COPII disassembly, and knockdown of PRRC1 can affect the efficiency of ER transport as well as membrane associations of COPII coats. The C-terminus(263-445) of PRRC1 contains a nucleotide phosphatase (NTPase) domain. To investigate whether the NTPase domain of PRRC1 is critical for its function, His-PRRC1(263-445)-HA has to be generated for carrying on further experiments.

CRISPR Cas9 Based Therapeutic Strategy for Alzheimer's Disease Supervisor: IP Nancy Y / LIFS Co-supervisor: FU Kit Yu / LIFS Student: WONG Lok Yin / IRE Course: UROP1100, Summer

Familial Alzheimer’s disease(FAD), which accounts for about 5% of all Alzheimer’s disease cases, is caused by one of more than 330 gain-of-function mutations in amyloid precursor protein (APP), presenilin 1 (PSEN1), and presenilin 2 (PSEN2) genes. Previous studies have shown that allele-specific disruption of mutated APP or PSEN1 by targeting mutations directly with CRISPR-Cas9 can effectively alleviate amyloid-related pathologies. Although this gives a promising cure to FAD, more than 330 CRISPR–Cas9 constructs are required for all identified FAD-related mutations, which is not feasible. This project aims to develop a generalized treatment for all FAD patients by targeting non-disease-causing single-nucleotide polymorphism(SNP). Here, this report systematically validated performance of sgRNAs for targeting candidate SNPs.

15

School of Science Division of Life Science

Deciphering Epigenetic Changes in Cancer -**UROP-EPIC program** Supervisor: LEUNG Danny Chi Yeu / LIFS Student: KANG Juwon / BIOT Course: UROP1100, Spring

The purpose of the UROP 1100 was to equip myself with techniques widely used in the field of molecular biology to proceed with coming up a feasible research question that could be explored for the FYP. The following content of the report introduces a detailed timeline and the learnings that I had over the UROP period.

Receptor Based Drug Development from Chinese Herbal Medicine Supervisor: LI Ning / LIFS Co-supervisor: XUE Hong / LIFS Student: HO Wing Chi / BIBU Course: UROP1100, Spring

Interactions of drugs with specific molecular targets allow them to alter the activities of these targets. Receptors have always been a major category of target and enabled the discovery of many drugs. Meanwhile, there is a revival of drug development from plants, due to issues such as severe side effects. Practised for thousands of years, Chinese herbal medicine presents a huge source of effective medicinal herbs and therefore plays an important role in this recent trend of drug development. This review will briefly present some recent progress in receptor-based drug development from Chinese herbal medicine in the novel COVID-19 and diabetes. Although there are hurdles to be overcome, Chinese herbal medicine still possesses enormous potential in receptor-based drug development.

Receptor Based Drug Development from Chinese Herbal Medicine Supervisor: LI Ning / LIFS Co-supervisor: XUE Hong / LIFS Student: KOT Yung Kei / BIBU Course: UROP1100, Summer

Anxiety is one of the most common mental issues around the world. It is much related to the benzodiazepine (BZD) binding site of the gamma-aminobutyric acid type A (GABAA) receptor. Therefore, it is indispensable to understand the principle behind and find possible method to minimize the impacts from it. In this review paper, the types of ligand-gated ions channels (LGICs), structure, functions and detailed inhibitory process of GABAA receptor and the introduction of BZD are to be discussed. Wogonin and baicalin are the flavones isolated from Scutellaria baicalensis Georgi (S. baicalensis) which show significant anxiolytic effect. They have valuable potential to be applied in the therapy in the upcoming future.

16

School of Science Division of Life Science

Receptor Based Drug Development from Chinese Herbal Medicine Supervisor: LI Ning / LIFS Co-supervisor: XUE Hong / LIFS Student: LAU Ming Ho / BIBU Course: UROP1100, Summer

Insomnia has always been a serious public health problem, affecting millions of people. Compared to common insomnia drugs like benzodiazepines, certain Chinese herbal medicine can manage insomnia with fewer side effects. The mechanism of action is mainly associated with the activation of GABAA receptors, which create sedative and hypnotic effects when the binding sites interact with agonists like GABA neurotransmitters. In this review paper, two kinds of GABA receptors will be introduced. Several sedative herbal medicines and their mechanism will be listed and explained. Some modern extraction procedures of herbal drugs like fractionation, chromatographic identification, and diverse identification will also be included.

Chinese Herbal Medicine for General Health Supervisor: LIANG Chun / LIFS Student: CHAN Yi Kai / BCB Course: UROP1100, Fall UROP2100, Spring

Inulin and stevioside are both acclaimed sugar substitutes in food industry since they manifest a healthier lifestyle to customers. Besides their competent sweetness, they have been reported to possess certain antidiabetic effects subject to the practical formulas of administration. And the single-ingredient antidiabetic effect of both inulin and stevioside were well-verified by my experiments conducted during UROP 1100. Besides the intriguing antidiabetic effects of these two natural sweeteners, their anti hyperlipidemic effects are also studied by several papers. (Huang, 2016) (Ahmad, 2018) Therefore, in these semester, non-alcoholic fatty liver HepG cells were exploited as a hyperlipidemic model undergoing regular treatments with different concentrations of inulin and stevioside to further investigate and characterise the anti-hyperlipidemic effects of inulin. Protein assay, TG measurement, western blotting, and fluorescence microscopy were utilized to provide the proof of concepts for a satisfactory performance of both inulin and stevioside in acting against hyperlipidaemia.

17

School of Science Division of Life Science

Chinese Herbal Medicine for General Health Supervisor: LIANG Chun / LIFS Student: KWON Daye / BIOT PARK Joo Hyoung / BIOT Course: UROP2100, Fall UROP1100, Fall

Traditional Chinese Medicine has been studied to have various applications in diseases and disorders, including that of the metabolic syndrome. In the present experiment, the efficiencies in the treatment of insulin resistance of three popular TCMs—Astragalus membranaceus, Poria cocos, Taraxacum were compared to that of a commonly known diabetes drug, Metformin. Insulin-resistant HepG2 cells were treated with each drug solution and were screened with addition of a glucose assay kit. Absorbance at 490 nm of the treated HepG2 was measured and a western blotting analysis was performed for quantification of Akt protein. Although Taraxacum showed the highest uptake of glucose, the results also suggested high efficiencies of glucose uptake stimulation and potentials of Astragalus membranaceus and Poria cocos.

Chinese Herbal Medicine for General Health Supervisor: LIANG Chun / LIFS Student: NG Cheuk Yan Gloria / SSCI Course: UROP1000, Summer

Diabetes is a major global medical issue that jeopardizes general public health. There is an increasing interest on the approach of Traditional Chinese Medicines (TCM) to complement diabetes treatment in Western medicines. The study aims to evaluate the effect of some TCM, Momordica Charantia (MC), Glycyrrhiza Uralensis (GU), and Ginseng (GS), on insulin resistance (IR) of HepG2 cells. The glucose uptake of drug-treated cells is measured to test the anti-insulin-resistant effect of drugs. The experiment results show that MC, GU, and GS exhibit an anti-insulin-resistant impact, where its effectiveness increases with increasing concentration and during drug treatment. In conclusion, using TCM as an anti-diabetic treatment can be a possible approach in Western therapies for diabetes.

Chinese Herbal Medicine for General Health Supervisor: LIANG Chun / LIFS Student: TSANG Chun Kit / SSCI Course: UROP1000, Summer

Certain herbal medicines have been used in Traditional Chinese medical practices to treat diabetic symptoms. Astragalus membranacues and Codonopsis pilosula are considered to possess hypoglycemic effects in a dose dependent manner 48 hours after adding to insulin resistance HepG2 cells. This effect does not manifest a dose dependent manner in the 24 hours trial which might be related to experimental flaws. The usage of TCMs demonstrates much smaller toxicity to HepG2 cells compared to typical anti-diabetic medicine metformin.

18

School of Science Division of Life Science

Chinese Herbal Medicine for General Health Supervisor: LIANG Chun / LIFS Student: WU Zehaoyu / BCB Course: UROP2100, Fall

Hyperlipidaemia has been stressed for its threat to the cardiovascular system in the recent years and people has been developing counteractions to the disorder. Traditional Chinese medicine (TCM) has also demonstrated its effects on the regulation of this disorder. This study utilized HepG2 cell line to establish a hyperlipidaemia cell model to examine the preventing and relieving efficacy on hyperlipidaemia of two TCM single herbs, Chen Pi and Huang Qi by examining their effects on AMPK through Western Blotting. With the observed lipid controlling effects, possible mechanism of lipid regulation of two tested TCM components was preliminarily established.

Anticancer Drug Resistant Mechanisms Supervisor: LIANG Chun / LIFS Student: AU Chun Kei / BISC Course: UROP1100, Fall

Cancer refers to uncontrollable cell division which can be characterized by multi-daughter cell division during the cell division cycle in our body. Multi-daughter cell division will give rise to more than 2 daughter cells which deviated from the normal cell division pathway and cycle. To investigate the effect of multi-daughter cell on cancer progression, we used 2 cell linages, 2-3 and 2-7. By making comparison with HeLa, we can figure out the cell length, multi-daughter cell division percentage and cell division duration to identify any abnormalities in cell division cycle of a cancer-resistant cell linage. The 2-3 and 2-7 cell linage was obtained from M2, an anticancer drug. By identifying the phenotypical appearance of the 2-3 and 2-7 cell linage, we can deduce their anticancer drug resistance behavior and their survival stress under progressive M2 drug concentration gradient.

Human Complex Disease Genomics and Bioinformatics Supervisor: LIANG Chun / LIFS Co-supervisor: XUE Hong / LIFS Student: HUI Cheuk Him / BIBU Course: UROP2100, Fall UROP3100, Spring

This study aims to identify the common SNV patterns for brain tumors, hence providing a genomic basis for early diagnosis and treatment. AluScan sequencing was performed on 10 sample sets of various types of brain tumors: mainly Meningiomas and Glioblastoma. SNVs were analyzed by comparing the genomic DNA from a brain tumor with white blood cells controls for each sample set. Upon comparison of mutation profiles across samples, different types of brain tumors were found to share similar SNV patterns as follows: The frequency of loss of heterozygosity (LOH) is markedly higher than that of gain of heterozygosity (GOH). GOH is enriched with NCG-to-NTG and ACT-to-AGT alterations, whereas LOH is enriched with NTG-to-NCG, ACC-to-AGC, and ACT-to-AGT alterations.

19

School of Science Division of Life Science

Human Complex Disease Genomics and Bioinformatics Supervisor: LIANG Chun / LIFS Co-supervisor: XUE Hong / LIFS Student: LIU Ran / BCB Course: UROP1100, Fall

Brain cancer is strongly associated with genetic variants. However, genetic variants in human are distributed in clusters and these clusters are known as genetic variant hotspots. Alu elements are transposons essentially contributes to these hotspots due to high-frequency recombination, mainly causing copy number variations. In this project, regions near and between Alu elements were prepared to sequenced using Alu PCR which used primers homologous with Alu elements to specifically locate regions near Alu elements. Blood samples from brain cancer patients are used. DNA from blood samples was extracted and amplified using Inter-Alu PCR. Then the products were fragmentated and selected to size 300 to 400 bp, and genomic AluScan DNA libraries were constructed. Analysis and evaluation of the process of each step were also performed in the project.

Human Complex Disease Genomics and Bioinformatics Supervisor: LIANG Chun / LIFS Co-supervisor: XUE Hong / LIFS Student: LIU Zhiyin / BCB Course: UROP1100, Spring

Cancer genomics has developed rapidly in recent years and has been applied to the clinical treatment of cancer. This report is a review on cancer genomics development and related techniques. Hallmarks of cancer have been summarized, providing a foundation for comprehending cancer biology. Researchers study several sorts of mutations that cause tumor formation and developed strategies and techniques to tackle these difficulties. Genome-wide association study (GWAS) and exome sequencing are applied to identify coding driver mutations. Whole genome sequencing (WGS) is utilized to detect non-coding driver mutations, and many models are used to correct the mutation rate during the procedure. Exome sequencing and singlecell RNA sequencing are used to investigate epigenomics, which, in addition to mutations, plays a crucial role in cancer genomics.

DNA Replication-initiation Proteins in Human Cells Supervisor: LIANG Chun / LIFS Student: SHI Chenwei / BCB Course: UROP1100, Spring

The mutation from normal cells to cancer cells is dependent on the abnormal DNA replication. Meanwhile, the disordered replication initiation is an essential part of the cancerization of normal cells. In this report, I review the literature that focuses on several DNA replication initiation proteins which are related to the cancerization of cells. Scientists have used many different methods to study the targeted proteins. For example, silence of the targeted mRNA by the oligodeoxynucleotides and siRNA molecules, the quantitative real-time PCR, CRISPR-Cas9 etc. So far, some proteins have been proved that they can be a marker that distinguish the normal cells and the cancer cells. Also, some proteins are potential target sites that helpful to treat cancer.

20

School of Science Division of Life Science

DNA Replication-initiation Proteins in Budding Yeast Supervisor: LIANG Chun / LIFS Student: TSUI Nok / BIBU ZHANG Cancan / SSCI Course: UROP1000, Summer UROP1000, Summer

The nucleolar complex-associated protein 3 (Noc3) is essential for the initiation of DNA replication. Understanding the role and function of this protein in DNA replication will lay the groundwork for future applications in cancer treatment. Noc3p is bound to chromatin throughout the cell cycle and plays an essential role in the formation of the pre-RC complex during G1 phase. A temperature sensitive mutant yeast strain, noc3-9, was transformed with a plasmid expressing FLAG-Orc2 and Orc2-Myc. Co- Immunoprecipitation Assays (Co-IP) were performed using G1-phase cells expressing the aforementioned proteins. The results indicate that at the permissive temperature, positive co-IP between FLAG-Orc2 and Orc2-Myc was possible. Contrastingly, at the non-permissive temperature the co-IP results were negative. The data suggest Noc3p is required for ORC dimerization in G1-phase.

Molecular Regulation of Axon Regeneration Supervisor: LIU Kai / LIFS Student: GAO Zhouyang / BCB Course: UROP4100, Fall

Every year, more than 250,000 patients are facing the difficulties from spinal cord injury. Unfortunately, for now there is no effective treatment to the spinal cord injury. Though function recovery can be achieved in injury to peripheral nervous system of human, injury on spinal cord, a component of the central nervous system, faces failure of axon regeneration. Therefore, understanding the difference between PNS and CNS axon regeneration can provide important insight to the treatment permanent disabilities from spina cord injury. Previously, we have used DRG cell as a model to study the intrinsic mechanism of axon regeneration. Currently, we are looking into PC12 cells as another effective system to facilitate the process of finding novel gens involved in axon regeneration.

Molecular Regulation of Axon Regeneration Supervisor: LIU Kai / LIFS Student: HARTANTO KWEE Jeffrey / SENG Course: UROP1000, Summer

Nerves in the peripheral nervous system (PNS) are capable of axon regeneration, but those in the central nervous system (CNS) are not. It is thought that the ability of axons to regenerate declines throughout development. Gene expression patterns changes from development to adulthood, and scientists are trying to identify the genes that inhibit or promote axon regeneration. Scientists study nerves that can grow back in all kinds of species, and through overexpression or inhibition of genes, try to enable axon regeneration in the CNS. Axon regeneration is currently possible in vitro, and scientists are working to ensure regeneration is quick and leads to functional recovery. This report aims to summarize some background information on axon regeneration and presents the results of some of the experiments that I have done.

21

School of Science Division of Life Science 22 Molecular Regulation of Axon Regeneration Supervisor: LIU Kai / LIFS Student: LIU Jingyi / BIBU Course: UROP1100, Fall

The adult mammalian central nervous system's (CNS) failure to regenerate axons is linked to two features of the CNS: a restrictive extrinsic environment and a decreased intrinsic regenerative capacity of mature CNS neurons. The goal of the research is to find and analyze new genes involved in axon regeneration, particularly those that are regulated intrinsically. Inability of the injured axon to regenerate and rebuild their function is a fatal problem after central nervous system injury. Even if they sometimes regenerate, it is almost impossible for them to connect with its original target. In contrast, strong regeneration occurs after peripheral nerve injury, which can lead to the recovery of sensory and motor functions.

Molecular Regulation of Axon Regeneration Supervisor: LIU Kai / LIFS Student: TAN Lianxing / IRE Course: UROP1000, Summer

PC12 cell line derived from a pheochromocytoma of the rat adrenal medulla. It is one of the most used cell lines in neuroscience research, such as neuron secretion, toxicity, and axon regeneration due to its neuronal properties and ease of culture. We, therefore, tested a series of compounds using a PC12 cell line to mimic the effects of these compounds on normal neurons. And it was found that CP-105696, a Leukotriene B4 receptor (LTB4R) antagonist, significantly impacted PC12 neurite growth. By looking for downstream signals, several possibilities for this effect are suggested.

The Rules of Packaging Fat in Cells Supervisor: MAK Ho Yi / LIFS Student: CHIU Chung Wah / SSCI Course: UROP1100, Summer

Triacylglycerols (TAGs) are the main stores of metabolic energy and fatty acids in human body. Abnormalities in TAG biosynthesis could disrupt the lipid homeostasis. Two acyl-CoA diacylglycerol acyltransferase (DGAT) enzymes, DGAT1 and DGAT2 are identified to account for the vast majority of TAG biosynthesis in human. Although the C. elegans DGAT2 ortholog has been identified, the identity of the DGAT1 ortholog is unknown. Here we investigate the effects on C. elegans intestinal lipid droplet (LD) phenotype upon the deletion of a putative DGAT1 ortholog in C. elegans.

22

School of Science Division of Life Science

Modeling Peroxisomal Disorders with C. Elegans Supervisor: MAK Ho Yi / LIFS Student: FUNG Ching Nam / SSCI Course: UROP1000, Summer

Peroxisome plays an important role in human metabolism, including b-oxidation of very long chain fatty acid and synthesis of plasmalogen (Wang & Subramani, 2017). Mutations in PEX genes, which encode peroxins, could result in peroxisome biogenesis disorders such as Zellweger spectrum disorder (ZSD) (Wang & Subramani, 2017). It was previously reported that peroxisomal beta-oxidation defects in C. elegans caused abnormal lipid droplet expansion (Zhang et al, 2010). Results here shows the effect of diet intervention on lipid droplets size and peroxisomal matrix protein transport in prx-10 mutated C. elegans. Additional plasmids were constructed for further interrogation of PRX-10 function in C. elegans in the future.

miRNA Processing Supervisor: NGUYEN Tuan Anh / LIFS Student: JUNG Kwangsek / BIOT Course: UROP1100, Summer

With the growing interest in microRNA's role in cellular functions, its biogenesis and maturation are being studied more intensely. This project looks into the effects of mutations in the sequence of the micro-RNA 142 and how it influences the maturation step from primary micro-RNA to pre-micro-RNA. 3 separate mutations, previously reported in human cancer genome sequencing have been opted for analysis. MiRNAs were synthesized in vitro and were used for cleavage assay with Drosha3-DGCR8 microprocessor complex. This report has demonstrated how mutations in the different region affect the cleavage efficiency and the position of the cleavage.

Targeting Mitotic Regulators in Cancer Cells for Potential Treatment Supervisor: POON Randy Yat Choi / LIFS Student: CHING Chi Yin / BCB Course: UROP2100, Fall UROP3100, Spring

It has been found by our laboratory that during mitosis, Programmed death-ligand 1 (PD-L1) expresses a second band (other than the normal band found in interphase) in Western blots. We hypothesize that PDL1 could be post-translationally modified during mitosis. We intend to investigate the identity of the second band by transfecting various modified constructs of PD-L1 before inducing a mitotic block. It is proven that the transfected PD-L1 shows a similar pattern as the endogenous one which is mentioned in the previous report. For these studies, we focus on cloning and transfecting truncated PD-L1 to study the position of the modification. The studies indicate that the 133-290 truncated and 1-132 truncated PD-L1 only show a single band pattern and with no mitotic shift in the western blot when using HeLa as cell models.

23

School of Science Division of Life Science

Targeting Mitotic Regulators in Cancer Cells for Potential Treatment Supervisor: POON Randy Yat Choi / LIFS Student: CHONG Man Yan / IRE Course: UROP1000, Summer

Chinese hamster ovary (CHO) cells are an important host for therapeutic protein production. However, its potential as a gene knockout system for characterizing the role of cell cycle control genes have not been developed. Our laboratory has recently developed an approach for constructing conditional gene silencing cell lines in HeLa cells by integrating the transcriptional control of the tetracycline-controlled promoter system and the post-translational regulation of auxin-inducible degron (AID). In this study, we investigated the possibility of constructing a similar system in CHO cell lines by transfecting a plant F-box protein (AFB2) and tetracycline-controlled transcriptional activator (tTA) genes into CHO-AA8 through the Sleeping Beauty transposon system. Moreover, we tested the possibility to arrest CHO cells in G1 phase with the drug lovastatin.

Targeting Mitotic Regulators in Cancer Cells for Potential Treatment Supervisor: POON Randy Yat Choi / LIFS Student: SINGHOF Michelle / BCB Course: UROP1100, Spring

Cell wounding refers to plasma membrane damage. Plasma membrane repair is integral to the survival of cells after wounding and involves a multi-step process. Ca2+-influx triggers the exocytosis of vesicles such as lysosomes. The extracellular release of lysosomal sphingomyelinase then triggers caveolar-endocytosis of the plasma membrane lesions. This process is distinct from conventional clathrin-mediated endocytosis. Cell wounding is especially common in mechanically active tissues such as skeletal muscle, and plasma membrane repair defects at any of these stages have been linked to a number of myopathies. Identification of actors in plasma membrane repair thus may be key to finding targets for new therapies.

Targeting Mitotic Regulators in Cancer Cells for Potential Treatment Supervisor: POON Randy Yat Choi / LIFS Student: SUN Run / BCB Course: UROP1100, Fall

Apoptosis is the process of programmed cell death which is critical for cells to survive. Normal cells use this mechanism to clean the body of cells that are irreparable to prevent cancer cells. However, abnormal cells such as cancer cells have deficiency in this function and may grow without control. So, it is indispensable to study the mechanism behind the process of apoptosis and unveil the essential genes related to it. In this study, we try to uncover the relationship between some relevant genes and MARCH5, a known E3 ubiquitin-protein ligase localized in the mitochondrial outer membrane, and the pathway of how the deficiency of MARCH5 can lead to mitotic cell death in cancer cells. The novel gene UBE2J2 is reported to have high correlation with MARCH5. We found that UBE2J2 shows discrepancy in different tissues. However, the knockout of UBE2J2 was not very successful in the first part of the experiment. Also, direct interaction between MARCH5 and UBE2J2 was also not shown via immunoprecipitation.

24

School of Science Division of Life Science

Targeting Mitotic Regulators in Cancer Cells for Potential Treatment Supervisor: POON Randy Yat Choi / LIFS Student: THEN Claudia Regina / BCB Course: UROP1100, Spring

Chinese Hamster Ovary (CHO) cell is frequently used for the production of recombinant protein. There are several steps to creating a culture of CHO cells expressing recombinant protein. First, the gene of interest is amplified using polymerase chain reaction (PCR). Then, the amplified DNA fragment and recipient plasmid DNA are digested using restriction enzymes to create complementary ends. After that, the digested DNA fragments are isolated using gel purification. The gene of interest and plasmid DNA are then connected using ligation procedure to make a complete plasmid. Lastly, the plasmid is transfected into CHO cell culture. This report will describe in more detail the standardized version for each procedure as the experiments have not yet been performed in actuality.

CRISPR/Cas9 Analysis of Essential Genes Supervisor: POON Randy Yat Choi / LIFS Student: BYUN Seung Min / BIBU Course: UROP1100, Fall

DHX9 is an ATP-dependent RNA helicase that may function in the cell cycle. I conducted a project on transfecting DHX9 CRISPR into a HeLa cell to confirm its effect of knockout as well as its potential function in the cell cycle. Using calcium phosphate-mediated transfection and Western blotting, I showed that the CRISPR was not effective in knocking out the gene of interest. Additional constructs of DHX9 CRISPR were designed for future experiments.

CRISPR/Cas9 Analysis of Essential Genes Supervisor: POON Randy Yat Choi / LIFS Student: KIM Jimin / BIOT Course: UROP1100, Spring UROP2100, Summer

Cells can be synchronized at specific phases through different ways such as chemical blockade, serum deprivation, centrifugal elutriation, etc. Among, RPE1 cells are increasingly widely being used as models in biomedical research. However, compared to HeLa, which is the most commonly used human cell line, an effective way to synchronize RPE1 cells has not been studied yet. Therefore, this report aims to examine a way to achieve cell synchronization of RPE1 cells by attempting different methods, including serum starvation and the utilization of Lovastatin. After conducting FACS with the samples, at which phase of the cell cycle the cells were trapped was explored.

25

School of Science Division of Life Science

CRISPR/Cas9 Analysis of Essential Genes Supervisor: POON Randy Yat Choi / LIFS Student: LIN Muyun / BCB Course: UROP1100, Summer

In vivo gene knockout, like RNA interference and the CRISPR/Cas-9 system are robust and powerful tools in research studies and have been proved to have a wide variety of applications. Here, I tested the effectiveness of a new technique involving nanobodies which aims at degrading proteins post-transcriptionally via the highly conserved ubiquitin pathway. As the exploratory stage of investigating the functions of nanobodies, starting with the relatively well-studied protein, GFP (Green Fluorescence Protein), will stand as an excellent example for later studies. In total, three different nanobodies were tested: the NSlmb-vhhGFP4, AIDvhhGFP4 and SPOP-vhhGFP4. Preliminary results showed that the nanobodies worked in degrading H2B-GFP in a considerable effectiveness.

Cloning and Characterization of Novel Cytoskeletal Regulators Supervisor: QI Robert Zhong / LIFS Student: CATIELLO Carlo / BIOT Course: UROP1100, Fall UROP2100, Spring

Microtubules are major components of the cytoskeleton which are formed by α and β tubulins and are essential for the organization of eukaryotic cells, playing an important part in vital cellular processes such as intracellular transport, structural support, and cell division. The initiation of microtubule growth, microtubule nucleation, occurs at the microtubule-organizing centers (MTOCs) and is regulated by the γ- tubulin ring complex (γTuRC), which mainly consists of γ-tubulin and γ-tubulin complex proteins (GCPs). From the previous experiment, it was found that a phosphomimetic mutant of γ-tubulin, γ-tubulin-GFP-Twinstrep S364D have binding defects to GCP proteins, making them harder to be incorporated into y-TuRC. We will conduct experiments, RNA interference knockdown as well as microtubule regrowth assay to further study how mutations affect nucleation in centrosomes.

26

School of Science Division of Life Science

Cloning and Characterization of Novel Cytoskeletal Regulators Supervisor: QI Robert Zhong / LIFS Student: TSE Kai Lee / BICH Course: UROP1100, Fall UROP2100, Spring

Microtubules are the cytoskeleton in eukaryotic cells that consists of α- and ß-tubulin heterodimers. They are highly conserved among species and have noticeable inherent dynamic properties. Random point mutation of human cells using CRISPR technology followed by screening could be used to discover new regulator proteins of microtubules nucleation. In this experiment, Fluorescence-Activated Cell Sorting (FACS) is used to carry out microtubule regrowth assay in mutated retinal pigmental epithelial cells (RPE-1). This report continued the UROP1100 project, which focuses on optimization of microtubule regrowth assay in suspension cells before passing through FACs, aiming to minimize other factors that may affect the signal strength. TX-100 and Tween-20 are used as detergents to permeabilize the cell. Methanol fixation was introduced and tested individually or in combination of glutaraldehyde fixation to study their effect on cell membrane permeabilization.

Standardization of Traditional Chinese Medicines Supervisor: TSIM Karl Wah Keung / LIFS Student: CHOI You Jin / BIOT Course: UROP2100, Fall

Cholinergic system contributes significantly to signal transmission in the brain and the muscle. Many literature reviews and studies propose that the presence of cholinergic system also play a crucial role in nonneural functions in human epidermis, by regulating essential activities in keratinocytes and melanocytes. The expression of cholinergic proteins, the release of acetylcholine, and the regulation of Alpha7 Nicotinic Acetylchonine Receptor (α7 nAChR) have been determined by many studies; however, the mechanism and the influence of cholinergic system on the dermatological functions have not been clearly identified. A more comprehensive understanding on the physiology and mechanism of cholinergic system would provide a more elucidated explanation, prompting a further development on therapeutic solutions and novel treatments for various diseases.

27

School of Science Division of Life Science

Standardization of Traditional Chinese Medicines Supervisor: TSIM Karl Wah Keung / LIFS Student: KIM Seo Hyun / BIOT Course: UROP1100, Fall UROP2100, Spring

First found in 2019, Covid-19 has caused severe casualties and impact on every segment of the world leading to economical slump. To overcome the pandemic, scientists have been developing drugs to treat the disease to promote active social interactions without the risk of infection and death. Traditional Chinese Medicine has been put on the spotlight for a potential resolution for effective treatment. Although several drugs had been developed and some are still under standardization, TCMs such as Jinhua Qinggan granules, Tetradrine, Cepharanthine and Nelfinavir have proven of their efficacy through multiple studies. With multiple herbs with different cellular mechanism involved, they are able to alleviate pulmonary disease symptoms from mild (cough, headache) to severe (fluid retention in lungs).

Standardization of Traditional Chinese Medicines Supervisor: TSIM Karl Wah Keung / LIFS Student: NG Siu Fung / BCB Course: UROP1100, Spring

Danggui Buxue Tang (DBT) is a traditional herbal formula that is proven to be capable of alleviating menopausal symptoms. Premature ovarian failure (POF), the abnormally early occurrence of menopause, is not a well-studied disease and thus there is currently no effective treatment to recover ovarian function. Two recent studies found that DBT might be able to treat POF by suppressing apoptosis. Quercetin and kaempferol were found to be bioactive components that could exert a certain effect similar to DBT on rat models. These two compounds could act as markers for the standardisation of DBT. Combining in vivo experiments and in silico analysis, not only could the therapeutic effect of DBT but also the pathogenesis of POF be elucidated.

Computational Study of Long Noncoding RNAs in Cancer Supervisor: WANG Jiguang / LIFS Student: WU Yurong / CHEM Course: UROP4100, Fall UROP4100, Spring

Cancer, which is an incurable disease, is one of the most popular topics among biologists. Despite this, the problem remains unsolved. In this project, we approach the problem using long-noncoding RNAs (lncRNAs), which were considered not as functional as protein-coding genes because it does not encode mRNAs that translate to proteins. Later, more research found that some processes are related to lncRNAs. With RNA sequencing (RNA-Seq), we can obtain the expression data of various lncRNA genes in different samples of cancers. Analyzing them can help us find out the pattern of expression of these genes in various types of cancers, which helps elucidate the mechanisms and develop treatments and drugs that cure cancers.

28

School of Science Division of Life Science

Imaging Genomics: Developing Machine Learning Methods for Precision Cancer Diagnosis Supervisor: WANG Jiguang / LIFS Student: LIU Zhiyin / BCB Course: UROP1100, Summer

The handy tool Connectivity Map (CMap) links genes, small molecules, and disease states through shared gene expression signatures. It is commonly utilized in producing and analyzing perturbational datasets to advance the identification of innovative medicines and help us understand human disease. Since the small amount of the dataset limits its capabilities, researchers enlarged the data volume tremendously in the next generation Connectivity Map with L1000 high-throughput gene expression profiling technology. A specific type of cerebral cavernous malformation (CMM) is defined by somatic MAP3K3 mutation in our previous research and thirty up-regulated genes were founded and used in the next generation Connectivity Map by L1000 for analysis and identification of candidate drugs for CMM.

The Application of BIG DATA Technologies in Precision Cancer Medicine Supervisor: WANG Jiguang / LIFS Student: LEUNG Kung Nam / BTGBM Course: UROP1000, Summer

With the lifetime risk of cancer uprising worldwide (Ahmad, A. S., Ormiston-smith, N. & Sasieni, P. D., 2015), cancer prognosis, or the evaluation of cancer development and survival analysis, is playing a more significant role in making treatment decisions. While clinical factors for prognosis include age, race, ethnicity, tumor histological type, and therapeutic history, this project aims to access the feasibility of incorporating clinical features alongside pan-cancer features, which are the data on frequently mutated genes and other genomic abnormalities (Illumina, Inc., 2022), for survival time prediction using machine learning methods. The project focuses on Glioblastoma multiforme (GBM) and carries out a fundamental analysis of the provided GBM dataset, followed by an evaluation of different regression techniques in survival prediction. The goal is to provide a foundation for further development of more advanced models on the dataset, including deep learning models.

Study of Blood Cell Development Using Zebrafish Model Supervisor: WEN Zilong / LIFS Student: FAN Yining / BCB Course: UROP3100, Fall

mpeg1.1 has long been used as a marker gene for macrophages in mouse and zebrafish. However, recently there are increasing evidence of the heterogeneity of mpeg1.1+ cells in adult zebrafish. Therefore, to give a throughout classification of mpeg1.1+ cells, we sorted DsRed+ cells in epidermis, gut, gill, kidney and brain of adult Tg(mpeg1.1:loxP-DsRed-loxP-GFP) zebrafish using FACS (Fluorescence Activated Cell Sorting) and conducted single cell RNA sequencing. Six distinct cell population was isolated and further experiments were conducted to verify the distribution of these cell populations and explore the functions of each cell population.

29

School of Science Division of Life Science

Study of Blood Cell Development Using Zebrafish Model Supervisor: WEN Zilong / LIFS Student: LI Zhuying / BCB Course: UROP2100, Fall UROP3100, Spring UROP4100, Summer

γδ T cells are considered as a T cell subset that play an intermediate role between innate and adaptive immunity. They display T cell receptors (TCRs), but have distinct way of recognizing their target antigens. They do not recognize peptide antigens presented by antigen presenting cells, but directly respond to infection consequence, giving them more innate-like features. In zebrafish, γδ T cells mainly distribute in the skins, gills and intestines. chemokine (C-C motif) ligand 38, duplicate 6 (ccl38.6) is a gene exclusively exists in zebrafish and may be closely associated with γδ T cells. This report will introduce the progress of our project on γδ T cells via research on ccl38.6.

Calcareous Dinoflagellates Supervisor: WONG Joseph Tin Yum / LIFS Student: MONG Sze Hang / BIOT Course: UROP1100, Fall

Due to elevation of carbon dioxide concentration in the atmosphere, the pH value of seawater has been decreasing over the decades, which may facilitate the calcite dissolution and fluctuated the bicarbonate and carbonate content in seawater. Given the fact that carbonate ion is one of the building blocks of biomineralization of calcareous dinoflagellates, it is proposed that ocean acidification will affect the calcification and growth of calcareous dinoflagellates. Having established a fundamental knowledge regarding the various life cycle alterations of Thoracosphaera heimii. from the previous project, this report would like to investigate how increasing bicarbonate and carbonate ion will affect the morphology as well as the cell population of T. heimii.

G Protein-coupled Receptors in Cell Signaling Supervisor: WONG Yung Hou / LIFS Student: LEE Tung Yeung / BCB Course: UROP1100, Spring UROP2100, Summer

Guanine nucleotide-binding protein alpha-subunit-i (Gαi) is known to interact with adenylate cyclase (AC) and exert an inhibitory effect on the conversion from ATP into an important secondary messenger cAMP. However, the domain(s) contributing to its inhibitory effect is not completely understood.

Together with the results obtained from the last semester, we generated 8 sets of Gαi1 mutants at the chi1 (residue 297-318) region aiming at mapping out all the residues that may interact with AC. The chi1 region, together with N256, K257, and W258 in Gαi1, had been shown to effectively reduce the inhibitory effect on AC. Moreover, we have confirmed the protein expression levels of the mutants generated for future functional studies.

30

School of Science Division of Life Science

G Protein-coupled Receptors in Cell Signaling Supervisor: WONG Yung Hou / LIFS Student: LIU Zhiyin / BCB Course: UROP2100, Fall

Melatonin receptors (MT1 and MT2 in mammals) are G protein-coupled receptors (GPCR) which play an important in signal cascade of melatonin. As mentioned in the last research report, we applied Tango assay to determine the dimerization of MT1 and MT2 and firstly attempted to construct plasmids expressing melatonin receptors and TEV protease but failed. Therefore, the purpose of the research this semester was to find out the key factors for successful transformation and improve the transformation efficiency. The results showed that the ratio of vector and insert as well as plasmid concentration were imperative for transformation efficiency, by which we can optimize the original method.

G Protein-coupled Receptors in Cell Signaling Supervisor: WONG Yung Hou / LIFS Student: LO Vincent / SSCI Course: UROP1100, Summer

Nm23-H1 (Non-metastatic protein 23 H1) is known to be a metastasis suppressor for most cancers. Meanwhile, SUSD2 (Sushi domain-containing 2) can be both a metastasis suppressor and promoter concomitantly. Recent research shows they may have a unique relationship inside the cancer cells. Thus, we measured the protein level of Nm23-H1 and SUSD2 in both breast and lung cancer cell lines by immunoblotting in this project. Based on the result, we proposed that PI3K/Akt and Wnt signaling pathways would be the feasible interaction pathways between two proteins in different cancer cells. Additionally, miRNA and galectin-1 may also involve in their connection. Therefore, we will work on their protein-protein interaction in the future. We believe finding out their association will advance the current anticancer drugs and therapy.

G Protein-coupled Receptors in Cell Signaling Supervisor: WONG Yung Hou / LIFS Student: YAU Hiu Laam / BCB Course: UROP1100, Fall UROP2100, Spring

Intense research has been engaging in the mechanism of the Gα-mediated downstream signalling pathway since the discovery of GPCR signalling in the 1990s. As a component of heterotrimeric G protein, the Gβγ subunit that is dissociated from the Gα subunit upon receptor activation also plays a crucial role in regulating downstream effectors in GPCR signalling. Due to the redundancy of Gβγ and Gα signalling pathways, the effector interacting capability of Gα must be first removed so that the Gβγ-mediated signalling response induced by receptor activation can be observed in vitro. In this study, we attempted to identify the critical residues or motifs responsible for the effector interaction in Gαi1 by studying the crystal structures of Gαi1 and adenylyl cyclase 5.

31

School of Science Division of Life Science

Big Data: Bioinformatic Analysis of Single-cell Genomic Data Supervisor: WU Angela Ruohao / LIFS Student: SAPUTRA Alexander William / SENG Course: UROP1000, Summer

The need to process high-throughput sequencing data is growing in conjunction with the rapid adoption of next-generation sequencing technologies. One method of transcriptome assembly is de novo assembly, where reads are assembled into transcripts without the use of a reference genome. This report discusses a popular software used for this purpose, called Trinity, including the algorithms behind it, evaluation of its performance with respect to other de novo assemblers (Price and IVA) and a demonstration of a typical workflow centered around Trinity (assembly and annotation).

Big Data: Bioinformatic Analysis of Single-cell Genomic Data Supervisor: WU Angela Ruohao / LIFS Student: WU Ka Hei / GBUS Course: UROP2100, Fall

As the global population continues to age, age-related neurodegenerative disorders are predicted to increase to 135 million cases worldwide by 2050, overtaking cancer as the second leading cause of death after cardiovascular disease (Gammon, 2014). In this study, we aim to create a single-nucleus RNA seq (snRNA-seq) library to facilitate the study of neurodegenerative disorders. We focus on analyzing the gene expression in three regions of the mouse brain – cortex, hippocampus, and cerebellum and identify any aging-related molecular and cellular changes in the brain by comparing mice of 3-months vs. 17-months of age. This progress report will focus on comparing the changes in cell composition within the cerebellum and their relevance in the development of neurodegenerative disorders.

Big Data: Bioinformatic Analysis of Single-cell Genomic Data Supervisor: WU Angela Ruohao / LIFS Student: YU Jiamu / SSCI Course: UROP1100, Summer

Bioinformatics, an emerging scientific subdiscipline where biological data is integrated and interpreted by demanding computational and statistical methods, has always been an intriguing topic. Harnessing the power of algorithms and data science, this interdisciplinary subject is empowered to uncover a substantial world beyond biology, fulfilling the scientific understanding of biologic processes, specifically COVID-19 pathogenesis. This report records an application of single-cell analysis on a nasal swab from a paediatric patient infected with SARS-CoV-2. By visualising and annotating the nasal swab dataset, this report attempts to attribute to the standard transcriptome database of COVID-19 for future studies, which provides hints for potential clinical and pharmaceutical applications.

32

School of Science Division of Life Science

Interdisciplinary Bioengineering: Microfluidic Device for Cell Manipulation/Culture and Molecular Biology Supervisor: WU Angela Ruohao / LIFS Student: PARK Junhyeok / BIEN Course: UROP1100, Spring

DNA sequencing is one of the most important techniques in science and medicine, as it can tell a lot of information about DNA sequence of molecules, such as cancer cells or virus. Knowing them is beneficial in our lives as they can be used to figure out treatments to diseases as well. This report will explain what DNA sequencing is, then explain what the different types of DNA sequencing are including traditional Sanger sequencing as well as Next-Generation Sequencing (NGS) such as Illumina. It will then illustrate what are different methods in preparing DNA fragments that can be used in DNA sequencing. Furthermore, some of applications of DNA sequencing will also be addressed, which include how DNA sequencing helps treatment of cancer in patients. Finally, some data analysis will also be addressed.

Genetic Studies on Stem Cell Regulation Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: HO Lok Yin / BCB Course: UROP1000, Summer

Adult stem cells are unspecialized cells that can generate and replenish a tissue. Its abilities of self-renewal and differentiation combined with gene editing will provide the potential for regenerative medicine (Kanherkar et al., 2014). Given that stem cells still have various uncertainties regarding their usage and safety, further study on stem cell biology is required. Analysis of Germline stem cell (GSC) development in Drosophila gives insights into the regulatory mechanism. Recent studies indicate that multiple signaling pathways and interactions with the niche are involved in the progeny differentiation of GSCs (Tu et al., 2020). Genetic screenings using tissue-specific Gal4-UAS are conducted for identifying genes that regulate self-renewal and differentiation in Drosophila GSCs. Until the 18th of August, 3 genes showed abnormal phenotypes in cystoblasts (CB). In future studies, further investigation on phosphate and kinase genes will be conducted.

Genetic Studies on Stem Cell Regulation Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: HO Wing Chi / BIBU Course: UROP1100, Summer

Stem cells possess great clinical potential in areas such as regenerative medicine due to their ability to self-renew and differentiate into one or more cell types. Thus, understanding stem cells is essential to fully harness their potential. Germline stem cell (GSC) is a type of adult stem cell which is responsible for gamete production. Studies on GSC development have always been a crucial aspect of advancing our knowledge about stem cells. Drosophila is one of the effective models for studying GSC behavior. In this study, Drosophila ovaries are examined, and genetic screening is performed to identify genes involved in GSC development. Seven genes are hitherto found to be related to GSC development. Their specific roles in GSC development will be further investigated in future studies.

33

School of Science Division of Life Science

Genetic Studies on Stem Cell Regulation Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: LEE Man Ping / BIOT Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

Kinases take an important role in the phosphorylation processes in vivo in order to maintain normal cell functioning. Some cyclin-dependent kinases are being identified to be involved in the self-renewal or differentiation of germline stem cells in Drosophila. However, the function of many kinases in Drosophila remains unknown. Here, I have conducted a genetic screening for 50 transgenic RNAi genes encoding for kinases to determine their roles in GSC self-renewal and differentiation. 23 genes out of the 50 genes investigated show a reduction in the number of germline stem cells, in which the knockdown of CG17528, cerk, cdk9 and CG11811 show more significant phenotypes. In future studies, I will further investigate the interaction mechanisms of the kinases identified in the germline stem cell niche.

Genetic Studies on Stem Cell Regulation Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: LEUNG Sin Ho Henry / BIBU Course: UROP1100, Fall UROP2100, Spring

Stem cell research has been started to carry out worldwide. Utilizing stem cells could cure severe diseases. Stem cell therapy such as introducing genetically modified hematopoietic stem cells into a patient’s bone marrow could cure leukemia. Induced pluripotent stem cells are found to turn an adult cell back into a stem cell, increasing the flexibility of the usage of stem cells. Drosophila serves as a common model for scientists to discover the mechanisms within a different types of stem cells. That knowledge may be useful in understanding certain similar genes related to stem cells in humans. Therefore, this report includes a summary of stem cell research genetic strategies in Drosophila.

34

School of Science Division of Life Science

Genetic Studies on Stem Cell Regulation Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: LIN Muyun / BCB Course: UROP1100, Fall

The term ‘transposable element’ (TE) refers to the piece of DNA sequence that is capable of changing its own position within the genome by inserting it into random positions. In previous research it has been identified that piwi-interacting RNA (piRNA) acts as a master repressor in the silencing of TE, especially in Drosophila, bringing genetic stability as well as maintaining germline stem cells (GSCs). 5-methylcytosine (5mC) is an abundant RNA modification that controls many aspects of RNA metabolism, like RNA stability, splicing, nuclear export, and so on. However, it remains unknown if RNA 5mC modification regulated TE silencing. In this study, by using a TE/piRNA reporter, Burdock-GFP, we found that RNA 5mC modification is involved in TE regulation. Overexpression of a dominant-negative form of Yps (YpsW68A, has been shown to affect RNA 5mC modification in the previous study), and knockdown of RNA 5mC methyltransferases, Nsun2 and Nsun4, significantly increases Burdock-GFP expression, indicating TEs are upregulated when disrupting RNA 5mC modification. In the future study, we will focus on investigating the molecular and genetic mechanisms that how RNA 5mC modification controls TEs silencing.

Genetic Studies on Stem Cell Regulation Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: ONG Gavyn Kye Wen / BCB Course: UROP1000, Summer

Stem cells generate new copies of existing cell types for differentiation and replacement of old or damaged cells and produce more stem cells for infinite stem cell production of an organism. This could be investigated in Drosophila ovaries. Virgin female flies with nos-gal4 and c587-gal4 transgenes were crossed with male flies with 896 RNAi lines and one control line. The offspring were collected, taken care of and cultured for maturation. Mature female flies were dissected for its ovaries and immunostaining was performed. The ovaries were mounted and viewed under a confocal microscope. The abnormal phenotypes were captured and investigated in how these abnormalities affect stem cell regulation.

35

School of Science Division of Life Science

Genetic Studies on Stem Cell Regulation Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: TANG Xiaohan / BCB Course: UROP2100, Fall UROP3100, Spring UROP4100, Summer

The Drosophila ovary has been an elegant model for studying processing such as sexual reproduction, stem cell biology and developmental signaling. Decades of extensive research has successfully revealed the entire process of germline development, from stem cell niche maintainence to differentiation, cyst development and oocyte determination, these events occurring in the germarium pushes the formation of a premature egg.

Genetic Studies on Stem Cell Regulation Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: XU Rui / BCB Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

Previous studies have indicated that protein kinases are fundamental in many biochemical reactions. Therefore, nowadays, kinases have also become the critical sites of many researchers’ attention when studying the regulation of cell signaling. In this project, we aimed to investigate the role of different kinase coding genes in the cap cell and terminal filament mediating germline stem cell development in Drosophila by crossing Bab1-Gal4 with 30 kinase RNA interference (RNAi) lines (detailed could be found in the methodology part). We did a genetic screen for different kinds of protein kinases. Moreover, eventually, we discovered that CG6612 and CG82957, CG17146, and CG 4032 have stimulating effects on the suspension of germline stem cells (GSC).

Niche Regulation of Stem Cell Self-renewal, Differentiation and Ageing Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: PARK Changmo / BIOT Course: UROP1100, Spring UROP2100, Summer

The impact of a particular gene on stem cells could be investigated by knocking it out with RNA interference and studying for possible abnormal phenotypes or any differences compared to the phenotypes of wild types. If genes involved in germline stem cell development in drosophila are silenced, many defects could be found such as abnormal follicle structure in ovaries, deficiency in the number of GSCs due to self-renewal defects, and the unusual number of cytoblasts due to differentiation defects. Throughout this genetic screening research, several genes crossed by GAL4-NOS or GAL4-C587 were found to have such abnormal phenotypes especially the BL Flybase Stock number of 35372, 41931, and 41889. In future studies, the reason for their abnormalities will be studied and identified.

36

School of Science Division of Life Science 37 Niche Regulation of Stem Cell Self-renewal, Differentiation and Ageing Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: WONG Sze Chai Helen / IRE Course: UROP1100, Summer

The Drosophila ovary harbors a community of germline stem cells to produce oocytes for reproduction, yet some of the detailed genetic and molecular mechanisms remain unknown. Thus, a preliminary genetic screening is particularly effective in searching for genes controlling germline stem cell development. By using the Gal4-UAS system to carry out RNA interference and silence selected kinase genes, we successfully identify fj, for, fu, fwd, Galk and gish as genes involved in Drosophila ovarian germline stem cell development. In addition, fs(1)h and fray are discovered to regulate ovary development. Results of this genetic screening have paved the way for future studies on the molecular pathways regulating germline stem cell self-renewal and differentiation.

Niche Regulation of Stem Cell Self-renewal, Differentiation and Ageing Supervisor: XIE Ting / LIFS Co-supervisor: TU Renjun / LIFS Student: ZHAO Yicheng / CHEM Course: UROP1100, Fall

In the Drosophila ovary, a small number of cells can maintain their ability to self-renew and generate daughter cells that can further differentiate, and these cells are referred to as stem cells. Germline stem cell (GSC) is one of the best models to study the molecular and genetic mechanisms that how stem cell self-renewal and differentiation are controlled in vivo. They are located in a structure named germarium, which is at the tip of the ovary. GSCs can differentiate into oocytes and can be readily identified by observing their fusomes. Within GSCs, RNA binding proteins play a vital role in the cell’s gene expression, specifically affecting RNA processing, export, localization, and translation. In this project, we try to study the subcellular localization of some RNA-binding proteins in the Drosophila ovary, using the method of Drosophila ovary immunostaining. We found that Imp and Fmr1 are expressed in the cytoplasm, while CG5728 and nonA-I are localized in nuclear, indicating they play different roles in regulating germline development. In the future study, we will investigate which RNAs are associated with these RNA binding proteins, and the molecular functions that how these RNA binding proteins regulate RNA metabolism.

37

School of Science Department of Mathematics

Department of Mathematics Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: CHAU Yu Hei / DSCT Course: UROP1100, Spring UROP2100, Summer

This document is a progress report that indicates an attempt in this summer to generalize the result of Guan- Li which proves shows a variant of the Alexandrov- Fenchel inequality for star shaped domains. Some conditions are identified to yield results similar to Guan-Li's, however it would ultimately be shown that the conditions are not satisfied in some sense except for the known 𝜎𝜎𝑘𝑘 functions, so no new interesting results can be established.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: CHUNG Tsun Ho Anson / MATH-PMA Course: UROP2100, Fall UROP3100, Spring UROP1000, Summer

This progress report contains results derived during the last two months regarding general curve shortening flow. The first major result concerns when a translating soliton of the general curve shortening flow can be contained within a finite strip. This problem can be studied using elementary observations and ODE comparison principles. Essentially, we require the speed function to be bounded by the square root of curvature when curvature is close to 0. The second major result establishes a more general form of the Harnack inequality. Previous results considered only cases when the speed function is a monomial, or satisfies certain convexity requirements. Here we set out a condition that is satisfied by a large class of functions including polynomials and non-homogeneous functions.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: LAU Kwun Chai Michael / MATH-PM Course: UROP1100, Summer

This report is to give a short summary on smooth manifolds and the motivation behind for developing this mathematical object. For example, definitions and some of the important result about regular hypersurface, manifolds with or without boundary, tangent spaces of manifolds, orientability and submanifolds will be given. Finally, using the result from smooth manifolds, such as immersion theorem and submersion theorem, we can show that the intersection curve between a regular hypersurface and plane parallel to tangent vector of the surface at certain point, is locally smooth and in fact a submanifold of the hypersurface.

38

School of Science Department of Mathematics

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: SHIH Jackson / QFIN Course: UROP1100, Fall UROP2100, Spring

Continuing from the last semester, where we have acquired basic knowledge of Riemannian geometry, we started reading and have finished going through the classical paper in Ricci flow by Hamilton. After that, we started explore literature on our own to broaden our horizon and try to perhaps come up with some toy problems we could work on. As the starting point, I have scanned through a few papers on mean curvature flow. In the study of hypersurfaces evolving under (inverse) mean curvature flow, monotone quantities are often applied to proofs of some important results. In this report, I would first introduce the preliminary knowledge of geometry on hypersurfaces, mean curvature flow and present some results I have read through.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: SU Zhaohao / DSCT Course: UROP1100, Summer

Riemann manifold is one of the most important study objects in modern differential geometry and a basic topic in undergraduate mathematics study. In this report, we are going to review some interesting propositions and results during the study of calculus on manifold and Riemann geometry, such as the Cartan’s Magic formula, the local expression of the normal vector of a smooth hypersurface and relate it with the first fundamental form of Riemann manifold, and the orientability of real projective space with odd dimensions and try to give some brief proofs with basic computations to contribute to the better understanding of them.

39

School of Science Department of Mathematics

A Machine Learning Approach to Study the Relationship between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: CHEUNG Cheuk Lun Alan / MATH-AM Course: UROP2100, Fall

Gas chromatography-mass spectrometry (GC-MS) is one of the most commonly used methods to analyze the trace of organic pollutants in the air. The sample gas is passed through the capillary column Gas chromatography, and different components in the sample gas are separated. Other components leave the capillary at different times (retention times). Together with mass spectra, GC-MS shows the various components inside the sample gas and their abundance.

The main objective of this study is to develop computational pattern recognition algorithms for fast preliminary analysis for the identification of the components and their concentration in the sample gas. In this paper, an improved algorithm based on previous results will be introduced. Compared to the previous version, the decision tree method is added to work along with the known algorithm in identifying the peaks point of different components. Where the decision tree model divides the GC-MS graph into several parts, additional treatment is done for different parts to minimize errors. Also, some simpler but still efficient methods for finding peaks, starting points, and endpoints are introduced.

A Machine Learning Approach to Study the Relationship between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: EE Hong Zhi / SENG Course: UROP1000, Summer

The purpose of this project is to investigate the urban heat island phenomenon. Urban heat island is when temperatures in urban areas are higher relative to their unurbanized counterparts. This effect is seen in almost all heavily urbanized areas in the world. This project focuses on using machine learning techniques to confirm the positive correlation between urbanization and land surface temperature, specifically in the greater bay area. The data used in this project are land surface temperature, terrain, population density, and urban morphology data, the last of which is split into sub-datasets. After processing the data, it is used to train a deep neural network machine learning model, which was evaluated using mean absolute error as well as the correlation coefficient. The results support a positive correlation between urbanization and land surface temperature.

40

School of Science Department of Mathematics

A Machine Learning Approach to Study the Relationship between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: HE Qihao / DSCT Course: UROP2100, Fall

Neural Network (NN) and Random Forest Regressor (RFR) models are common tools in machine learning. Based on the result of our previous study, this study will examine the models’ performance in terms of accuracy and efficiency in predicting the Land Surface Temperature (LST) of the Hong Kong Islands. To improve the performance of the models, we designed a feature extraction method, tried a special way of train-test set splitting, and corrected a minor shift of the temperature image compared with other feature images using QGIS. The RFR has outperformed the NN with the optimal network structure among the tested configurations with two inner layers.

A Machine Learning Approach to Study the Relationship between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: MOK Wan Hin / MATH-IRE Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

In the previous semester, we built machine learning models that study the quantitative relationship between urban morphology and the heat island effect by relating the former with Land Surface Temperature (LST) by using MODIS LST products and Random Forest models. We found that the model had a high correlation (more than 0.9) between the predicted LST and the measured LST over the whole year, but some of this came from the seasonal variations that the model observed. For regional variations in LSTs due to urban morphology, we found that the performance varied from day to day. This semester, similar models were hoped to be explored for Sentinel images.

A Machine Learning Approach to Study the Relationship between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: SHEN Che / DSCT Course: UROP1100, Fall

The goal of this research is to study the urban heat island effect, which is a commonly observed meteorology phenomenon. Since the traditional three-dimensional model is computationally costly with many uncertainties, we attempted to build machine learning models for predicting the land surface temperature (LST) by some possible influencing factors, including average building height (ah), total building surface (lb), population, terrain height, tree height and impervious surface percentage of corresponding locations. We gather the data of the land surface temperature (LST) and predictors with different formats, transform them into Excel form and process them to make the sample for regression. After the sample is accomplished, we use it to train different regression models, evaluate their performance and analyze the possible causes of deviation.

41

School of Science Department of Mathematics

A Machine Learning Approach to Study the Relationship between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: WANG Yuanfu / SSCI Course: UROP1000, Summer

The urban heat island (UHI) effect raises concern as global warming accelerates and extreme weather frequently appears. Previous research has revealed the relationship between the UHI effect and urban morphology, while how they are related remains uncertain. In this study, the UHI effect is measured by land surface temperature (LST) in the Greater Bay Area. Then, several models of LST in Hong Kong, including the linear model, deep neural network (DNN) model, random forest (RF) model, and support vector machine (SVM) model, are built based on urban morphology data, along with other geographical factors. All of the prediction results have high correlation coefficients to true values. They show similar geographic patterns to the measured LST, indicating that urban morphology data can be used to model the urban heat island effect. The flaws in dealing with the data and the immature algorithm are considered responsible for the errors, and several potential optimizations are proposed.

A Machine Learning Approach to Study the Relationship between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: ZHANG Hanning / COSC Course: UROP1100, Fall

Urban information is very important to many scientific research areas. However, most specific information is not easy to acquire directly. Therefore, methods of getting such information indirectly are sometimes needed through remote sensing, machine learning, reverse engineering, etc. In this study, we are trying to retrieve one of the most important building information (the physical dimension) using a machine-learning algorithm to recognize buildings in different high-resolution images acquired from advanced online maps. And the result of the study will help to provide more information for the study of urban modeling.

42

School of Science Department of Mathematics

A Machine Learning Approach to Study the Relationship between Urban Morphology and Urban Heat Island Supervisor: FUNG Jimmy Chi Hung / MATH Student: ZENG Lingqi / DSCT Course: UROP1100, Spring

The Land Surface Temperature (LST) is the radiative skin temperature of the land surface derived from solar radiation. A better estimation of the LST could be useful for predicting the urban climate. A well-trained AI model could be used to predict the future urban heat environment based on different scenarios and provide guidance for mitigation strategies. In this study, four machine learning models: linear regression, multi-layer perceptron, random forest regressor, and epsilon-support vector regression, are trained with urban structure data to predict land surface temperature. Bivariate spline approximation is applied while retrieving the data. The results and performance are also compared with a set of reference training data from a previous study.

Retrieving of Urban Morphology from Satellite Images and Machine Learning Supervisor: FUNG Jimmy Chi Hung / MATH Student: KAN Wan On / DSCT Course: UROP1000, Summer

Urban morphology, the study of urban forms, is essential to environmental science and city planning because it focuses on the agents and processes responsible for urban transformation over time. This research aims to apply machine learning techniques to satellite images in order to estimate building height information with low cost and high efficiency. As this project progressed in the spring of 2022, we conducted research and developed multiple image segmentation machines and learning models. Currently, our efforts are focused on enhancing pioneer models for more accurate ground truth predictions and image processing techniques for extracting and manipulating more appropriate open-source data sets. These techniques, as well as their outcomes and obstacles, are discussed in detail in this report.

Retrieving of Urban Morphology from Satellite Images and Machine Learning Supervisor: FUNG Jimmy Chi Hung / MATH Student: NG Ka Ho / MATH-IRE Course: UROP1100, Summer

In this project, we aim to improve the prediction of building dimensions from satellite images by upgrading the UNet approach developed in the previous progress in Spring 2022. By experimenting with modern architecture, Swin-UNet, enhancing data augmentation, altering prediction goals, and running systematically organized trials on a greater scale, we achieved minimum validation categorical cross-entropy losses of 0.322 and 0.048 respectively, on the low-quality target dataset and the high-quality control dataset. This report first revisits the motivations for extracting building dimensions for urban morphology. With the project objective stated, we record the rationale to introduce the above additions to the model together with the experiment results and analysis, during which a small diversion to a literature review of Swin-UNet is included for completeness. The programming side and the optimization of experimental conditions are isolated and discussed.

43

School of Science Department of Mathematics

Retrieving of Urban Morphology from Satellite Images and Machine Learning Supervisor: FUNG Jimmy Chi Hung / MATH Student: YANG Yuang / DSCT Course: UROP2100, Fall

Urban morphology is of great interest to a wide range of fields, including environmental science and city planning. In this project, we aim to deploy machine learning methods and satellite images to allow the retrieval of building height information at a low cost. Following the progress on this project in the Summer of 2021, several things have been tried out to improve the performance of the earlier developed model. Those methods include generating a 3D view of the buildings to obtain more accurate ground truth, trying out alternative satellite images as input, and exploring open-source datasets that can be used in this project in the future. This report describes these methods in detail, along with the results and difficulties encountered.

Retrieving of Urban Morphology from Satellite Images and Machine Learning Supervisor: FUNG Jimmy Chi Hung / MATH Student: ZHAO Haoyu / DSCT Course: UROP1100, Fall UROP2100, Spring

Urban information plays an important role in various scientific fields. One of the key features of urban information is the dimension information of urban buildings. However, the acquisition of three-dimensional information tends to be difficult and expensive in the past. But with the online map engine and open-source street view images, high-resolution satellite images are available, which provides chances to obtain building information at a low cost. This report will discuss a specific deep-learning method, the U-Net neuron network, and use four satellite image data sets to train a machine-learning model for building detection. After training 20 epochs, the overall accuracy can reach up to 0.7. The feature of data sets and the techniques used in extracting features in images are also discussed.

Quantum Groups Supervisor: IP Ivan Chi Ho / MATH Student: CHAN Hong Ming / COSC Course: UROP2100, Fall

In paper given by V.G. Turaev, a more general type of link in-variant is constructed which can be reduced to Jones polynomial. This report will show how the reduction to Jones Polynomial is performed. In another two papers given by N.Reshetikhin and V.G. Turaev, further generalized the invariant to coloured ribbon graph in closed oriented connected 3-manifold. This can used to generate a 2-variable Jones-Conway polynomial and an infinite set of 1-variable reductions of the 2-variable Kauffman polynomial according to N.Reshetikhin and V.G. Turaev. In this report, we will show some interesting property of the invariant Pm in V.G. Turaev.

44

School of Science Department of Mathematics

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: CHAU Yu Foon Darin / MATH-IRE Course: UROP1100, Spring

We investigate the construction of cluster algebras over the classical SL(V)-invariant ring of tensors. Fomin and Pylyavskyy pioneered this study of these rings in the context of cluster algebras using Kuperberg’s web basis. We provide a simplified, explicit construction for the procedures described in the original article. We classify the cluster algebra types for n = 9, and prove a structure theorem relating. We also suggest some applications to conjectures in the original reference by Fomin and Pylyavskyy.

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: CHIU Yan Ho / IRE Course: UROP1100, Summer

In this report, we define a turn of a snake graph generated by an arc. The number of the perfect matchings of the snake graph gives an integer sequence of type which the even terms of it satisfy the mutation exchange relation . We also use the cluster expansion formula in terms of perfect matching to show that it has a linear recurrence relation by considering its snake graph. We could link up the exchange relation and the linear recurrence relation by the relationship between cluster variable from surface and its corresponding snake graph.

Quantum Groups Supervisor: IP Ivan Chi Ho / MATH Student: CHUNG Soobeom / PHYS Course: UROP2100, Fall

Since its discovery, the Jones polynomial of links was shown to be constructed using quantum groups by Turaev. It was also generalized to isotopy invariant of colored links and ribbon invariant in . Furthermore, Witten constructed a topological invariant of the closed 3-manifold using Chern-Simonseld theory and the mathematical counterpart was demonstrated by Reshetikhin and Turaev. For an oriented 3- manifold obtained by surgery on a link, they used quantum groups and modular Hopf algebra to produce the invariant. The result is a linear sum of the generalized Jones Polynomial. In this report, derivation and an example of Jones polynomial will be shown. Furthermore, invariant for different 3-manifolds will be calculated.

45

School of Science Department of Mathematics

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: FOO Peace / MATH-PM Course: UROP2100, Fall UROP3100, Summer

The first part of this brief overview covers the basis and elementary methods used in the study of quantum cluster algebras and surveys the parallels between classical and quantum forms of cluster varieties with concrete examples of the quantization process given in the form of quiver representations. This serves as a prelude to a discussion of the Quantum Lift Theorem examined in great detail by Goncharov and Shen as a synthesis of several different results relating to moduli spaces, cluster variety structures, and quantum cluster algebras.

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: HO Sui Kei / COSC Course: UROP2100, Spring

Plabic graphs and plabic networks are combinatorial objects that appears in a number of cluster algebra system. We start by reviewing the definitions and combinatorics of plabic graphs. Next, we review an application of plabic networks in representation of quantum group and compare with the quiver representation.

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: KWOK Kin Ming / MATH-IRE Course: UROP2100, Fall

Cluster algebra was introduced in S. Fomin and Zelevinsky, 2001 and Sergey Fomin and Zelevinsky, 2007, which also introduced the concepts of c-vectors, g-vectors and F-polynomials. Subsequently, from Nakanishi and Zelevinsky, 2011, c-vectors and g-vectors are collected as column vectors into matrices, known as Cmatrices and Gmatrices, allowing linear algebra method to be taken advantage of in the study of cluster algebra. Two main conjectures - the sign-coherent conjecture for C-matrices and the Laurent positivity conjecture has been proposed, which was proved in Gross, Hacking, Keel, and Kontsevich, 2016 using scattering diagrams and use of toric geometry. Subsequently, an alternative proof in Nakanishi, 2021, was presented using also scattering diagrams but without the usage of toric geometry. The result from Nakanishi, 2021 would be the main focus of this report.

46

School of Science Department of Mathematics

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: PHAN Nhat Duy / MATH-PMA Course: UROP1100, Summer

We study cluster algebra with principal coefficients that are associated to unpunctured surfaces. We recall the formulas in Musiker and Schiffler that used perfect matchings of the graph to give the specific formula for the Laurent polynomial expansion of cluster variables in these cluster algebras. We then apply that result to generalize Laurent polynomial expansions over an annulus and a pair of pants.

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: TAI Sung Chit / COSC Course: UROP1100, Summer

This study aims to study the properties of the snake graphs on the once-punctured torus, and an equivalence relation is defined on the set of snake graphs corresponding to the arcs of the torus. Consequently, a unique decomposition theorem is proposed, where it is argued that a snake graph can be uniquely decomposed into 2 smaller snake graphs. The decomposition ensures that there is a surjection between positive coprime pairs of integers and the Markov numbers.

However, to solve the Markov Unicity Conjecture (MUC), which states that every Markov triple has a unique maximal element, the surjection needs to be a bijection. Therefore, a recurrence relation is proposed for computing the number of perfect matchings on a given snake graph by defining a state space on the snake graph. A few examples are provided at the end of the paper as an attempt to demonstrate the use of the recurrence formula.

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: ZHANG Ke / MATH-PMA Course: UROP2100, Fall

In this article, we summarize the study of cluster algebra in UROP2100 during fall 2021. In this semester, we went through the artical article “Cluster Algebras and Scattering Diagrams” by Tomoki Nakanishi, where we learnt about the proof of sign-coherence conjecture related to scattering diagrams. We start with reviewing some definitions in cluster algebra which will be used in the later part. Next, we introduce the definition scattering diagram and wall-crossing automorphisms. At last, we relate scattering diagram with cluster algebra and give an example.

47

School of Science Department of Mathematics

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: ZOU Huaiyang / MATH-PMA Course: UROP2100, Fall

This report introduces some results of Nakanishi's “Cluster Patterns and Scattering Diagrams” [Nak21] about the proof of sign-coherence and Laurent positivity. The major ideas originate from the work of Gross, Hacking, Keel, and Kontsevich using scattering diagrams and toric geometry. The paper of Nakanishi translates some results into pure cluster algebra languages but gives new representations and logic of the proof. In this report, we introduce some of the results which are still based on scattering diagram method.

Quantum Groups Supervisor: IP Ivan Chi Ho / MATH Student: CHAN Hong Ming / COSC Course: UROP3100, Summer

In this report, an elementary way to construct Lusztig’s Basis from Poincar´e–Birkhoff–Witt(PBW) Basis in Quantum Group of A, D, E type pointed out by Peter Tingley in [2] is accessed. In section 1, definition of Quantum group, a set of automorphisms, bar involution and PBW basis will be introduced. In section 2, explicit calculation have been done on the well know A2 case. First, introducing Lusztig’s Basis in A2. Then, the transforming matrix between PBW Basis and Lusztig’s Basis is calculated for A2 case. And is expressed as combination Lusztig’s Basis. This expression is claim to be useful in constructing Lusztig’s Basis according to [2]. In section 3, some reviews and observations towards [2] will be given.

[2] Peter Tingley. Elementary construction of Lusztig’s canonical basis. 2016. doi: 10.48550/ARXIV.1602.04895. url: https://arxiv.org/abs/1602.04895.

Quantum Groups Supervisor: IP Ivan Chi Ho / MATH Student: CHUNG Soobeom / PHYS Course: UROP3100, Summer

In this paper, the nineteen vertex model and its associated R-matrix will be introduced. It is related to the square lattice model with a spin-1 representation on each edge obeying the ice rule. After that, by using the Boltzmann weights of the vertex model, FRT construction will be used to show the algebra associated with the six-vertex model first and subsequently the 19-vertex model. The result is that they share an identical algebraic structure. Using different representation and parameters, one will be able to produce a corresponding algebra much richer in terms of the structure and with many different applications.

48

School of Science Department of Mathematics

Quantum Groups Supervisor: IP Ivan Chi Ho / MATH Student: MAN Ryuichi / MATH-IRE Course: UROP4100, Spring

We study the actions of the generalized Casimir operators of the quantum groups on the family of parabolic positive representations introduced in [6]. Our main results (Theorems 3.1, 3.4 and 3.5) provide an explicit algorithm to compute the action of the Casimir operators on from their action on the positive representations, which can be computed using the formula in [4, Theorem 6.2]. In this report, we first provide the necessary background knowledge on the root systems and the quantum groups, and introduce the notations used. Followed by this is a summary of the key results obtained from our previous research. Then, we proceed to state and prove the algorithm to compute the Casimir action, and illustrate its use with an example.

[4] I. Ip, Positive Casimir and central characters of split real quantum groups, Communication in Mathematical Physics, 344(3), 857-888, 2016. arXiv: 1503.00543. [6] I. Ip, Parabolic positive representations of, preprint, 2020. arXiv: 2008.08589.

Integro-differential Equations: Theory and Applications Supervisor: JIN Tianling / MATH Student: LIU Jiaming / MATH-IRE Course: UROP1100, Spring

We study the fundamental structure of the Sobolev space so as to have a more rigorous understanding of the weak solution of the boundary value problem of second-order partial differential equations as well as the variational principle for the principal eigenvalue of symmetric elliptic operators in partial differential equations. In this study report, the basic structure of Sobolev space is introduced, including the weak derivatives of locally integrable functions, in which we highlight the similarities and differences compared with ordinary derivatives. Besides, the elementary properties of Sobolev space are presented. Supplementary concepts, such as mollifier, are introduced. Moreover, after understanding the fundamental structure of Sobolev space, we consider the explanation of the weak solution of boundary value problems and the eigenvalues of symmetric elliptic operators, based on the properties of Sobolev space, linear algebra and real analysis.

49

School of Science Department of Mathematics 50 Integro-differential Equations: Theory and Applications Supervisor: JIN Tianling / MATH Student: SZETO Chun / MATH-IRE Course: UROP1100, Spring

It is a well-known fact that any positive solution to the problem on is radially symmetric (D denotes the unit ball). This paper includes a brief review of its proof using moving planes. Curiously, the analogous conclusion for the fractional Laplacian is also true, in the sense that radial symmetry is also observed in positive solutions of the problem. Organizing arguments made by existing papers, the proof will be presented in details, and the differences between the local and non-local case will be discussed.

Efficient Algorithms for Visualizing Dynamical Systems Supervisor: LEUNG Shing Yu / MATH Student: CHAU Wai Ming / MATH-SFM Course: UROP2100, Fall UROP3100, Spring

We propose a k-means clustering approach for detecting the Lagrangian coherent structure (LCS) in the continuous dynamic system. The idea is to cluster the whole particle trajectory over a finite time interval for many different initial conditions and use the mean square error from each cluster to quantify the deformation for each cluster. The advantage of the proposed approach is that the method tries to use the whole particle’s trajectory to measure the rate of separation between the adjacent particles instead of only the final location of the particle. To further improve the computational efficiency, we develop an adaptive approach to construct different subsamples of the whole particle trajectory based on a finite time interval. We will apply our method to two examples, including bounded and unbounded dynamic systems.

Efficient Numerical Methods for Dynamic Interface Supervisor: LEUNG Shing Yu / MATH Student: LI Yuchen / SENG XU Minrui / SSCI Course: UROP1100, Summer

This article uses MATLAB to implement the discrete QR method and visualize Lyapunov exponents of a regular dynamic system. We begin with a theoretical analysis of the FTLE, the Lagrangian method, and the discrete QR method, respectively, followed by their implementations. Finally, we present one numerical example and a comparison between these two algorithms.

50

School of Science Department of Mathematics 51 Efficient Algorithms for Visualizing Dynamical Systems Supervisor: LEUNG Shing Yu / MATH Student: XU Zian / DSCT Course: UROP1100, Fall

In the original approach, we partition different initial conditions based on the positions of particles at a fixed final time tN as in finite-time Lyapunov exponent (FTLE). However, two trajectories may have a large FTLE at some time between t0 and tN, even if they have a small FTLE at the final time tN. To address this problem, we want to partition different initial conditions of particles based on the whole trajectories from the initial time t0 to the final time tN. And we try to do this by clustering the matrix formed by the positions at different times between t0 and tN using k-means.

Efficient Algorithms for Visualizing Dynamical Systems Supervisor: LEUNG Shing Yu / MATH Student: YAU Wing Yan / MATH-AM Course: UROP2100, Fall

We investigate a quantity to describe the flow motion, in the sense that how the infinitesimal portion (area for 2d flow, volume for 3d flow) changes over time, as well as testing the performance of two different Lagrangian approaches, that one of them approximate the final positions of grid points to obtain the Jacobian matrix, while another one keeps track on the deformation of infinitesimal portion to do so. Numerical examples of compressible and incompressible flows are presented, and their expected behaviour and realcase behaviour are discussed.

Efficient Algorithms for Visualizing Dynamical Systems Supervisor: LEUNG Shing Yu / MATH Student: YIP Kwun Chuen / MAEC Course: UROP1100, Fall

The Finite Time Lyapunov Exponent (FTLE) is a numerical tool for calculating the rate of separation of close initial values in a dynamic system. The report shows an implementation of the FTLE into macroeconomic fields to find a suitable domain of parameters for an economy that implements an integrable feedback expenditure policy in a financial crisis. The report will cover the model and the economic intuition behind it. The variable mainly used is the speed of response of government targeted demand towards the actual demand of the economy and the magnitude of the integrable policy. With these two variables, the report shows the output level and output growth of the economy within a decade of the policy implementation.

51

School of Science Department of Mathematics

Efficient Numerical Methods for Dynamic Interface Supervisor: LEUNG Shing Yu / MATH Student: ZHAO Yubo / SSCI Course: UROP1100, Summer

This paper will study and discuss several methods for testing and comparing the accuracy of symplectic integrators representing a Hamiltonian system. Symplectic integration of Hamiltonian dynamics is already a developed method, and high-order (say 4th or 5th) explicit methods are established for separable Hamiltonian with quadratic kinetic energy. However, for general, nonlinear Hamiltonians, no explicit representation is found. Since symplectic integrators are numerical methods specialized for the Hamilton system and computation is excessively time-consuming, we have to balance accuracy and computing speed. Starting from the framework, we investigate methods for Hamiltonian of different orders, then compute and investigate their accuracy.

Some Aspects of High-dimensional Probability Supervisor: WANG Ke / MATH Student: WU Zhiang / COSC Course: UROP1100, Summer

The theory of random graph is related to both combinatorics and probability theory. It describes certain probability distributions over a set of graphs. Erdös Erdos and Rényi Renyi proposed the first model of random graphs, which has numerous variations and applications in engineering. In this report, we study 𝔾(n,m) and 𝔾(n,p) models and the evolution of a random graph.

Some Aspects of High-dimensional Probability Supervisor: WANG Ke / MATH Student: XU Yitao / MATH-SF Course: UROP1100, Summer

The whole report introduces some basic knowledge about concentration inequalities. We first introduce some tail probabilities of a single random variable and the sum of independent random variables, such as Markov’s inequality and Chernoff bounds. Then we show the proof of sharper Hoeffding bounds for bounded random variables. Next is the introduction to a martingale-based method which can be used to get the concentration of functions of independent random variables. Lastly, we introduce the ∅-entropy and its application to concentration inequalities.

52

School of Science Department of Mathematics

Lattice-based Cryptography, Pseudorandom Functions and Algebraic Number Theory Supervisor: XIONG Maosheng / MATH Student: SAMIN Thanic Nur / MATH-IRE Course: UROP1000, Summer

In this report, we present a summary of some topics discussed in “An Invitation to Arithmetic Geometry”[Lor97] by Dino Lorenzini. In particular, we talk about integral closures of rings, specifically the ring of integers and polynomials. We also discuss the curve Zf (𝑘‾), which is the set of zeroes of a polynomial f in an algebraically closed field and how it relates to the set of maximal ideals of 𝑘‾ [x, y]/(f). We also discuss the concept of localization and local rings. Finally, we give a brief overview of how a post quantum safe algorithm was attacked.

Lattice-based Cryptography, Pseudorandom Functions and Algebraic Number Theory Supervisor: XIONG Maosheng / MATH Student: YUEN Theodore / MATH-PM Course: UROP1000, Summer

This report follows the main reference by Dino Lorenzini, by introducing the concept of integral closure of commutative rings in a finite field extension of their respective field of fraction, follow by an important characterisation of local Dedekind domain, method of determining the integral closeness of an integral extension to a Dedekind domain and how it relates to the property of unique factorisation of ideals. The case of 𝑘‾ [x; y]/(f) will be discussed in more detail, where we can gain insight to some relations between algebraic properties and geometric properties, and we will also use this example to demonstrate some properties of a Dedekind domain shown in this report.

Research in AI and Machine Learning Supervisor: ZHANG Tong / MATH Student: XIONG Yixin / MATH-SFM Course: UROP1100, Fall

This report introduces some basic settings of the M5 Machine Learning competition, suggests some approaches to handle the data that appeared in the competition, analyze the top solution result and its advantages, tries the Lightgbm model and gives some generalizations and insights to further research and industry applications.

This report is a study of the M5 Competition, mainly focusing on solving the problem of how to handle the M5 data set in a primary state. It contains a small proportion of realization and innovation, and more thoughts will be brought up in the future

53

School of Science Department of Mathematics

Research in AI and Machine Learning Supervisor: ZHANG Tong / MATH Student: XU Yanbo / COSC Course: UROP2100, Fall

The advancement in object detection has benefited many fields including autonomous driving, robotics, and so on. Among the detection problems, 3D detection has been extensively studied by researchers. However, most existing 3D object detection methods rely heavily on the massive amount of labeled data, which is hard to acquire. In the context of autonomous driving, the collection of unlabeled data, i.e., stereo data and video, is relatively easy. To leverage these data, we proposed a semi-supervised training framework with an objectlevel photometric loss which provides a supervision signal for the detection task by Structure From Motion. Extensive experiments on KITTI and nuScenes datasets show the improvement on 3d detection tasks using our method, compared with state-of-the-art approaches.

54

School of Science Department of Ocean Science

Department of Ocean Science Phytoplankton Biomass and Community Composition in Hong Kong Coastal Waters Supervisor: LIU Hongbin / OCES Student: LEUNG Chak On / BCB Course: UROP1100, Fall

Water quality is one of the main issues in marine fish culturing. A lot of factors can be associated the change in the water quality, such as temperature, the concentration of nutrients and pollutants, and ocean current. Therefore, a constant monitoring in water quality is necessary. HKUST has been in collaboration AFCD to measure the water samples in all HK marine culture zones. In this report, temperature, dissolved oxygen and chlorophyll a concentration were measured in August and September 2021, and the causes of their relationship and changes over time were investigated. It turned out that temperature was irrelevant to the change in water quality, but several factors were critical to it.

Study on the Adaptation and Physiological Responses of Synechococcus Isolates from Different Oceans Supervisor: LIU Hongbin / OCES Student: HUANG Jingxian / SSCI Course: UROP1000, Summer

Synechococcus are one of the most widespread photoautotrophs cyanobacteria in the marine environment, diverse clades of Synechococcus stand significant roles in the ocean ecosystem and marine primary production. Studying of their physiological responses and adaptation to variable environmental conditions may give clues of the principle that how do Synechococcus distribute in the global oceans and subsequently influence the biogeochemical cycle. In this study, we isolated eleven strains of Synechococcus from different regions of ocean and transferred the original sample to do the daily sampling under certain specific temperatures to gain the growth rate of them. The thermal responses of diverse clades of Synechococcus under different temperatures is the focus in this study.

Study on the Adaptation and Physiological Responses of Synechococcus Isolates from Different Oceans Supervisor: LIU Hongbin / OCES Student: LI Ching / BISC Course: UROP1000, Summer

Temperature is one of the most important parameters impacting the growth of cynobacteria which is responsible for transferring energy down the food chain and act as the role of carbon sink. In our experiment, we obtained 10 strains of Synechococcus isolated from different ocean. Attempting to find out their growth rate at different temperature. Our data show that there is positive relationship between the growth rate of Synechococcus and temperature. We predict Synechococcus will have wider distribution and higher abundance in the future under the influence of global warming and illustrate possible impacts from the growing Synechococcus such as increased primary activity and possible harm to the shelves water food web.

55

School of Science Department of Ocean Science

Chaotic Mixing Supervisor: MAK Julian / OCES Student: POON Yiu Cho Joe / SSCI Course: UROP1000, Summer

Aquatic locomotion is an important topic, studying it helps us to understand how objects are moved in liquids. Marine Animals like seahorses and jellyfish have their own unique moving pattern in the water. Jellyfish push off the water under their bells to provide movement while seahorses move by their flexible tail. What we do in this program is to find out the science behind it. In simple terms, It is about how things swim. In the report, I will examine how I started the research program on the study of aquatic locomotion and show some of my study result on it. At last, I will use python to do some simulations on the relationship between the liquid and moving object.

Characterising Coral Reef Function across Anthropogenic Gradients Supervisor: WYATT Alex / OCES Student: CHEUNG Chun Fei / OST Course: UROP2100, Fall

Coral reefs support many marine organisms and provide ecological services. Through stable isotope analysis of samples from different field sites, the marine food web of Hong Kong can be constructed, and anthropogenic impacts across the East-West gradient could be investigated. Sea cucumber stable isotope analysis data has shown that geographical factors have a larger impact than seasonal variation in terms of δ13C and δ15N, while the opposite is observed in δ34S. Variation in δ13C and δ15N is also small between muscle and outer wall tissues, while a larger variation is shown in δ34S. How stable isotope analysis of δ34S is used as a tool in ecological research will also be discussed below.

Characterising Coral Reef Function across Anthropogenic Gradients Supervisor: WYATT Alex / OCES Student: WONG Sung Lit / BISC Course: UROP1000, Summer

Increasing anthropogenic input has changed the global climate system rapidly over the year. Due to that, the drastic surge of heat content has been observed in the world’s oceans. Coral, as a marine organism, has to cope with sudden thermal stress inevitably. For coral to develop thermal resilience, the daily environmental temperature was regarded as the most impactful factor. Nonetheless, the related mechanism still remains unclear. Some studies believed the transcriptome of coral plays an important role in developing thermal resilience. Therefore, we would approach the question from transcriptomic aspect. Acropora digitifera were collected. Collected Acropora will be fragmented into nubbins and exposed to heat stress. Photosynthetic efficiency (Fv/Fm) of the nubbin would be measured and RNA would be extracted for assay at each time point. This study’s result is expected to provide insight into coral’s thermal resiling capability and function under current anthropogenic impacts.

56

School of Science Department of Physics

Department of Physics Fourier Optics Supervisor: ALTMAN Michael Scott / PHYS Student: NGUYEN Xuan Tan / PHYS-IRE Course: UROP1100, Fall

In earlier models of low energy electron microscopy (LEEM) image formation theory, the spread in the energy of the imaging electron beam was modelled as a Gaussian function to reflect the imperfect temporal coherence of electron sources in experiments. However, this simplification of the energy spread function may be unable to provide an accurate simulation of an image formed in LEEM, especially at the limits where effects caused by chromatic aberrations dominate those induced by wave aberrations. In this report, we will provide a method to calculate the image formed by a source of arbitrary energy spread function, based on the existing Gaussian model. Then, simulated images of the Gaussian spread and that of a cold field emission source are compared to examine the validity of the former one.

The Geometry and Physics of Wrinkling Supervisor: ALTMAN Michael Scott / PHYS Student: FONG Ching / SSCI Course: UROP1000, Summer

This article aims to study the effect on the geometry of uniaxial-constrained monolayer graphene when subjected to current flow upon a series of tabletop experiments. It is observed that the potential gradient of the graphene is non-uniform over the region where it is suspended. To explain this, we proposed the geometry is non-Euclidean over the region, and it is governed by both the uniaxial-constraint and minimization of the energy of the sp2 bonded carbon by the Keating model. In other words, by minimizing the energy functional, one may obtain the geometry of the suspended monolayer graphene via optimization methods.

57

School of Science Department of Physics 58 Materials Characterisation of Superconducting Thin Films for Quantum Computing Applications Characterization Supervisor: JAECK Berthold / PHYS Student: LEUNG Ka Wun Casey / PHYS Course: UROP1100, Fall

Superconducting qubits is one of the most promising approaches for quantum computing and they are based on fabrications of superconducting thin films on microwave circuits. To improve the performance of the qubits, the thin film is optimized by reducing the oxides and changing the superconducting materials. In this project, we investigate three superconducting thin films, namely aluminium thin film, tantalum thin film, and High Entropy Alloys thin film. The Tantalum thin film is fabricated through RF sputtering and high entropy alloys are fabricated through DC sputtering. The goal of this project is to characterise the properties of various superconducting thin films. To characterise the thin film, we use difference microscopic and spectroscopy techniques such as X-ray diffraction, Atomic Force Microscopy, and X-ray photoelectron spectroscopy (XPS). This allow us to characterize the metal and the oxides on the surface, which is the key of microwave loss of qubit resonator. Characterize

Epitaxial Growth of Magnetic Kagome Metals Supervisor: JAECK Berthold / PHYS Student: LIN Yi-hsin / PHYS-IRE Course: UROP1100, Summer

Kagome metals have been identified as a promising class of materials hosting rich phenomenon, including superconductivity and topological flat band. However, most investigations were done on bulk crystal due to the difficulty of growing thin film directly on substrate. Introducing platinum buffer layer in between can facilitate the uniform atomic layer growth of two-dimensional Kagome metals, paving the way for exploring more novel physical phenomenon. This report introduces the process of growing platinum buffer layer from substrate preparation, along with various techniques and instruments required for material characterization.

Electromagnetic and Acoustic Metasurfaces Supervisor: LI Jensen Tsan Hang / PHYS Student: CHEUNG Ka Ho / SSCI Course: UROP1000, Summer

Metamaterials have unique electromagnetic properties unlike conventional materials occur in nature due to their artificially designed volumetric structures. Metamaterials gave rise to many and on-going researches such as invisibility, superlenses, medical devices, etc. However, the production of the nano-fabrication process on metamaterials is expensive and complicated. Hence, metasurface, a two-dimensional thin layer nano-structured material is introduced as a cheaper and simpler alternative approach to optical experiments. Here, we investigate the components and the coincident data in a two photons imaging setup. By studying the whole process flow of the experimental setup, we can have a clearer objective and acknowledgement of the reason behind the components of the setup as well.

58

School of Science Department of Physics

Electromagnetic and Acoustic Metasurfaces Supervisor: LI Jensen Tsan Hang / PHYS Student: LI Yui To / PHYS Course: UROP1000, Summer

Metasurfaces are planar structures with artificial designed nano-slots arrays which allow control on a range of properties of light. This feature enables metasurfaces to be applied as various optical components, and it can be used for imaging. In this work, we use two-photon interference to image an object with unknown polarization properties on the metasurface, and retrieve the Jones matrix elements of this unknown object by performing coincident measurements between the object and the references metasurface, whose polarization information are known. The retrieved profiles of the Jones matrix elements can then be used to construct an image of the object by mapping them to HSB color scale.

Electromagnetic and Acoustic Metasurfaces Supervisor: LI Jensen Tsan Hang / PHYS Student: SHEK Wing Him / SSCI Course: UROP1000, Summer

Metasurface, within their subwavelength thick layer, can bring an abrupt change to incident light. They enable new kinds of optical manipulation while exhibiting unique properties that are unprecedented in traditional optical components. Recently, due to their ability in manipulating light, there is growing interest in putting them into quantum imaging systems, these systems utilize quantum properties of light to produce better images. The first half of this article will give a brief review of quantum imaging and metasurface. The remaining half will introduce a project that integrates metasurface into quantum imaging to extract the transmission property of a test object, then will discuss some simulations done on the metasurface used as the test object, and lastly, my conclusion on the simulation results.

Monte Carlo Simulation of 2D Supramolecular Assembly Supervisor: LIN Nian / PHYS Student: HUI Lik Hang Alpha / SSCI Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

This is a report on simulation with supramolecular assembly using Kinetic Monte Carlo (KMC) with the use of Fortran. We will briefly talk about the coding of the logic behind the diagonal bonding and method for counting sizes of hexagonal structures. Then, we would have some concise simulations analysis regarding triangular molecules and atoms with different basic parameters and temperature cycle, with our main goal of forming the largest hexagonal structure.

59

School of Science Department of Physics

Monte Carlo Simulation of 2D Supramolecular Assembly Supervisor: LIN Nian / PHYS Student: WENG Chun Yu / IRE Course: UROP1000, Summer

In this report, the molecular assembly of v-shape molecule on a 2D plane is simulated using Kinetic Monte Carlo method to produce hexagonal structures. First, the theory of the simulation and the master equation of the algorithm would be briefly discussed. Then, the practical choices of equations used in program would be explained. After discussing the theory, several important simulation results would be shown to study the effects of different physical parameters on the yield, including diffusion energy, bond energy, rotational bond energy, temperature, and the concentration of molecules. Finally, the simulation data would be analysed to provide some generalisations. Analyzed generalizations

Organic Molecular Beam Deposition in Ultra-high Vacuum Supervisor: LIN Nian / PHYS Student: KURNIAWAN Michael Alexander / PHYS Course: UROP1100, Spring

With the emergence of various novel properties such as tunability of band gaps of 2D Metal Organic Frameworks (2D-MOFs), studying its band structure will aid the development for their usage in modern technology. This work thus attempts to simulate the band structure and the total density of states of 2D-MOFs with Honeycomb-Kagome lattice arrangement using MATLAB program. Band gap calculation are performed with varying values of potential and two different radiuses of the metal atoms which shows larger gaps with increasing potential while there is an unexpected decrease of band gap in certain values of range with the atom having radius of 0.2 nm which could stem from the affected organic ligand potential with the larger metal atom’s radius.

Organic Molecular Beam Deposition in Ultra-high Vacuum Supervisor: LIN Nian / PHYS Student: LI Hengjian / SSCI ZHANG Xiangxi / IRE Course: UROP1000, Summer UROP1100, Summer

The project aims to discover the minimum evaporation temperature for two types of molecules: Fe-PC and (NH2)8Ni-PC. The experiment method is organic molecular-beam epitaxy which takes place in the vacuum produced by two vacuum pumps. Molecules were heated to different temperatures to generate a molecular beam. The melting point of the Fe-Pc molecule is between 250°C and 270°C; the intact molecule can be successfully deposited on the substrate plate. But the (NH2)8Ni-PC molecule cannot deposit successfully. In the future, new components can be added to the chamber to enhance the efficiency and accuracy, and other molecules can also be studied using this method.

60

School of Science Department of Physics

Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: CHUI Shek Hei / IRE Course: UROP1100, Summer

The use of supramolecular nanostructures have been increasingly common in modern day technology, such as the improvement of drug efficacy, various electronic devices to the construction of modern semiconductors. As a result, more ways have been developed to aid the studying of its band structure, ranging from Quantum Monte Carlo methods, Density functional theory, scattering matrix method to free electron approximation. The objective of this project is to perform simulations on its band structures of different lattices, such as the Kagome and Lieb lattice, through adjusting different parameters in the MATLAB program. Trends relating different parameters were observed and the reasons behind were analyzed.

Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: FAN Ka Cheuk / IRE Course: UROP1000, Summer

Since the discovery of unique physical properties of band structure of 2D lattice, more researches focus on studying band structure of 2D materials. In this work, a 2D physical system of lieb lattice is simulated using MATLAB program. Through solving 2-dimensional Schrödinger Equation numerically in program, band structure and total density of states (TDOS) are plotted. By varying different parameters (including lattice constant, well potential and potential well radius), it shows that how the band structure and band width relate to different parameters.

Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: WONG Yip Chun / IRE Course: UROP1000, Summer

In this report, the physical system of kagome and lieb lattice formed by two-dimensional supramolecular nanomolecule is simulated, by numerically solving the 2-dimensional Schrödinger equation. This is performed by the construction of lattice geometry via potential wells or pillars in MATLAB, which then calculates the variables of the structure such as band structure, the total density of states (TDOS) and the local density of states (LDOS). The purpose of this research is to explore how different geometries and parameters affect the lattice’s physical properties and enhance the characters of the kagome and lieb lattice’s band structure.

61

School of Science Department of Physics

Theoretical Study of Properties of Real Materials Supervisor: LIU Junwei / PHYS Student: KWOK Yuk Lam / PHYS-IRE Course: UROP2100, Fall

In order to obtain information about crystal lattices, such as band structure and density of wave function, a Python package is built. The final target is that it works for arbitrary crystal lattices the user wish to understand. This report describe the outline of the codes and provides further explanations for important parts and concepts. We have also built a self-checking function that could verify our calculated Hamiltonian in the k-space from symmetries. To demonstrate our results, we present two examples in 2-dimension and 3- dimension respectively. Finally, as this is an ongoing project, our future directions would be discussed.

Theoretical Study of Properties of Real Materials Supervisor: LIU Junwei / PHYS Student: LIU Yidai / PHYS-IRE Course: UROP1100, Spring

Graphene is a strictly 2D crystal exhibiting many outstanding physical properties. Though having a short history, it has already shown great potential in both theoretical study and application. In this report, the energy bands of graphene are calculated using the tight-binding model, the existence of Dirac cones is derived using different methods, the energy bands are classified into pi and sigma bands, and the characteristic features such as the merging of bands at symmetry points are studied form the symmetry point of view.

Theoretical Study of Properties of Real Materials Supervisor: LIU Junwei / PHYS Student: NGUYEN Xuan Tan / PHYS-IRE Course: UROP1100, Spring UROP2100, Summer

The crystalline structure of CrSb in the NiAs phase exhibits interesting magnetic properties as the spins of Cr atoms lies along the c axis, with ferromagnetic (FM) couplings between Cr atoms within the hexagonal basal plane and antiferromagnetic (AFM) couplings between adjacent planes. The layered antiferromagnetic order of this material, in addition to its large spin-orbit coupling and high Néel temperature, can empower possible future applications in AFM spintronics. To understand the electronic and spintronic properties of CrSb, an eight-band tight-binding model is formulated and the electronic band structure is analysed using the spatial symmetry group of the crystal. D6h4

62

School of Science Department of Physics 63 First-principles Calculations for Real Materials Supervisor: LIU Junwei / PHYS Student: HEUNG Tsz Fung / IRE Course: UROP1100, Summer

Tight Binding model is a modelling method for calculating electronic band structure of crystalline solids. It assumes that electrons can only hop from one atom to nearest-neighbor (NN) or next-nearest-neighbor (NNN), making the calculation simpler. This assumption can be justified when the Coulumb potential of the solid is strong. In this report, the band structures of graphene and graphene nanoribbon will be discussed.

Self-learning Monte Carlo with Deep Neural Networks Supervisor: LIU Junwei / PHYS Student: AU Kei Yin Eltis / PHYS Course: UROP1100, Summer

Monte Carlo Simulation is a commonly used method to simulate different physical systems. In short, the theory is basically first to generate a set of possible outcomes or states underlining the law of physics, and those states can be further calculated as the physical quantities. In the simulation of the Ising model, those states will be the different spin configurations of a square lattice. With the given configuration, Thermodynamical quantities such as magnetization, energy, heat capacity, and susceptivity can be calculated. This paper has briefly review the different updating methods of the Markov chain Monte Carlo (Hastings, 1970) on the two-body spin system, which include the local update(Metropolis, Glauber)(Metropolis, Rosenbluth, Rosenbluth, Teller, & Teller, 1953), Global update(Wolff)(Wolff, 1989), and especially on Machine learning update(self-learning) (J. Liu, Qi, Meng, & Fu, 2017). The performance of different updates under the critical temperature has been discussed.

Observational Phenomenology of Supermassive Black Hole Supervisor: LIU Tao / PHYS Student: TRAN Duc Huy / PHYS-IRE Course: UROP1100, Summer

The recent success of the Event Horizon Telescope Collaboration (EHT) in imaging the supermassive black holes M87* and Sgr A* has opened up a new channel to extract information about Fundamental Physics. The image of a black hole observed on Earth, which follows the laws of General Relativity at zeroth order, may receive corrections from the physics phenomena beyond Standard Model. In this UROP project, we investigate the corrections to the image of a black hole that are induced by an axion cloud surrounding that black hole. Understanding these effects could lead to a potential method to detect axion, which is a popular dark matter candidate, using the images of supermassive black holes.

63

School of Science Department of Physics

Scanning Tunneling Probe Experiments on Materials with Novel Electronic Properties Supervisor: LORTZ Rolf Walter / PHYS Student: LAW Chak Nam / SSCI Course: UROP1100, Summer

This project aims to test a new home-made scanning tunnelling spectroscopy (STS) probe on a superconductor. We begin with explaining the principles of Andreev reflection point contact spectroscopy and tunnelling spectroscopy. Then, the principle of the probe is detailed, and the sample preparation process is explained. We explain the procedures of performing point contact spectroscopy at cryogenic temperature and finally analyse the experimental data. We will discuss the experimental problems and limitation we encountered and point out ways to improve the experiment.

Investigation of CP Symmetry in Neutrino Oscillation Supervisor: LUK Kam Biu / PHYS Student: CHAN Chung Yuen / PHYS FONG Shing Hei / PHYS Course: UROP1000, Summer

This project is to determine the charge-parity(CP) violating phase in the process that a muon neutrino turns into an electron neutrino. In weak interaction, CP symmetry can break, which means that the particle and anti-particle do not behave the same way. In this measurement, muon neutrino and anti-neutrino beams are generated through pion decay. When the muon neutrinos travel, some of them change to electron neutrinos. A 20000-ton detector filled with liquid scintillator placed at 100 km from the neutrino source is used to capture the neutrinos. The expected number of neutrino interactions is the product of neutrino interaction cross-section, appearance probability, number of muon neutrinos that are generated and number of atoms in the detector. The CP-violating phase can be determined by counting the number of detected electron-neutrino events and their energy distribution.

Charge Generation and Carrier Diffusion in Organic Solar Cells Supervisor: WONG Kam Sing / PHYS Student: CHANG Fan Ut / SUSEE Course: UROP1100, Fall

Non-fullerene acceptors (NFAs) are a popular research topic in the organic solar cells (OSCs) recently since they have shown several advantages including higher tunabilities and better spectral absorption in certain region over traditional fullerene acceptors (FAs). In this study, three different NFA-OSC devices, PM6:Y6, PM6:m-BTP-phC6 and PM6:m-BTP-phC6:Y6 are tested by measuring their transient photocurrent and transient photovoltage lifetimes. The results showed that PM6:m-BTP-phC6 has the longest TPV lifetime, indicating the slowest recombination of charge carriers; while PM6:m-BTP-phC6:Y6 has the shortest TPC lifetime, indicating the fastest charge extraction. Overall, PM6:m-BTP-phC6:Y6 shows the highest external quantum efficiency and the highest power conversion efficiency of 18.2%, showing that the ternary strategy of tuning the morphology of the NFA-OSCs has a great potential for further studies in the future.

64

School of Science Department of Physics

Neural Information Processing Supervisor: WONG Michael Kwok Yee / PHYS Student: HO Man Fung / PHYS-IRE Course: UROP3100, Fall UROP4100, Spring

In last semester’s report, a new method of calculating the past history inputs was introduced. However, upon further inspection, it was discovered that this method may cause information leakage, which where unobtainable future data is used for the prediction of current result. Therefore, amendments of the calculation method will be introduced in this report to avoid this problem. In addition, using the feature importance score, the best sets of inputs to be used in the prediction were constructed. Using the newly constructed input set, it was found that accuracy of prediction for multiple machines are better than single machine by about 10% on average.

Neural Information Processing Supervisor: WONG Michael Kwok Yee / PHYS Student: LAO Zhiquan / PHYS-IRE Course: UROP1100, Fall UROP2100, Spring

The traffic data of Taiwan Highway System collected by the electronic toll sensors are studied. Traffic latency of the highway system is studied through XGBoost. Previous UROP studies have separated the data into multiple clusters through Gaussian Mixture Model (GMM) and predicted the clusters separately. However, the axes of GMM still need to be refined. Two methods, namely the mutual information analysis and single input GMM, are tried to figure out the axes in this study. Although the former generally gives a better result, the latter one makes a large improvement in the Wednesday’s data.

Diffusion of Nanoparticles in a Potential Energy Landscape Supervisor: WONG Michael Kwok Yee / PHYS Student: CHEUNG Ho Tin / SSCI Course: UROP1000, Summer

The diffusion of nanoparticles in a potential landscape is important in biophysics. However, the diffusion of nanoparticles in a potential energy landscape is different from the normal diffusion as the particles and we do not know much about the properties of the diffusion in a potential energy landscape. With the mathematical model calculated from the Fokker-Plank equation by Wong, we could approximate the diffusion through a slope by dividing the slope into number of steps. We would like to investigate the effects of changing different parameters, including the temperature, the source location, the observation point, the potential difference of the landscape, etc., and shapes of the landscape on the time-varying transmission probability, with aids of the approximation, using Matplotlib in Python.

65

School of Science Department of Physics

Mechanics of Soft Composite Materials Supervisor: XU Qin / PHYS Student: YANG Chun Hin / IRE Course: UROP1100, Summer

Soft composite has shown a variety of properties by adding different liquid inclusion. The elastic response of the soft composite is also changed by the liquid inclusion. The classical Eshelby’s theory describes how does the modulus of the composite changes with different volume fraction. Recently, stiffening effect is found to occurred to a soft composite with liquid inclusion. The counterintuitive phenomenon cannot be explained by the Eshelby’s theory. In this paper, sample with different volume fraction are fabricated to investigate the elastic response of soft composite. The result shows that Eshelby’s theory can be used to explain the response for low volume fraction but fail in high volume fraction.

Quantum Computing Based on NV Center in Diamond Supervisor: YANG Sen / PHYS Student: FEI Zijian / SSCI Course: UROP1100, Summer

In this project, we are supposed to use a room temperature experimental setup to detect the charge state of the NV color center. The NV center is a kind of crystal defect in diamond formed by two lattice defects, a nitrogen atom and a vacancy. The NV color center can be in NV0 or NV - states, and NV - is used for quantum calculation. Therefore, in order to improve the quantum fidelity, we need to make a measurement and selection, and select the one only in NV - state. This is done by illuminating it with a 593nm laser and detecting the fluorescence intensity. We need to tune the 593nm laser, use AOM (acousto-optic modulator) to switch the laser on and off, and record the fluorescence count. We also need to optimize the parameters in our work.

Femtosecond Laser Pulse Induced Air-plasma Filament for Generating THz Radiation: Experiment and Simulation Supervisor: ZHANG Jingdi / PHYS Student: HO Ka Ying / MAE Course: UROP1100, Fall

The generation of T-rays is of much research interest because of its numerous applications in biomedical imaging, sensing and plasma diagnostics. The far-infrared radiation can penetrate non-polar materials such as paper, clothing, ceramics etc., and shows great potential in historical research.

Current high-energy THz generation relies on large accelerators, synchrotrons and free electron lasers, which are costly and difficult to access. With growing demand for THz generation, using tabletop ultrafast lasers for THz generation is of much research interest. One of the methods is femtosecond terawatt laser system, and T-rays can be generated via two-color laser mixing in gases. This report will briefly summarize the two proposed models explaining the phenomenon of two-color laser mixing, and typical experimental setups of THz generation.

66

School of Science Department of Physics

Femtosecond Laser Pulse Induced Air-plasma Filament for Generating THz Radiation: Experiment and Simulation

Supervisor: ZHANG Jingdi / PHYS Student: LIN Yuchen / COSC Course: UROP3100, Fall

Following the previous UROP 2100 research, a series of numerical simulations on the terahertz generation induced by the Orbital Angular Momentum (OAM) modulated light and TEMlm wave have been performed for UROP 3100. Because the original two dimensional numerical model is inadequate when applying to systems with no rotational symmetry, a new three dimensional simulation program has been developed and tested this semester.

Meanwhile, the assembly of the femtosecond laser pulse induced terahertz generation system and the alignment of optical instruments were also carried out. In the coming winter period, experiments will be performed to generate the terahertz wave which serve as a prerequisite for the investigation of terahertz generated by polarized light, and we aim to verify our proposed model clarifying the dynamics of terahertz generation in future experiments.

Femtosecond Laser Pulse Induced Air-plasma Filament for Generating THz Radiation: Experiment and Simulation Supervisor: ZHANG Jingdi / PHYS Student: PENG Huaiyue / PHYS Course: UROP1100, Spring UROP2100, Summer

Lorentz oscillator is a classical model to describe the motion of the phonons when a material is polarized by the electric field brought by the input light wave, where a guassian pulse is adopted. In this program, we stimulate the nonlinear polarization process in SiC via FDTD algorithm inMATLAB. The modulated equations are derived from the Lagrangian by assuming the varying permittivity and effective charge. After getting the dispersion spectrum and compare it with the classical one, we find that dispersion of the nonlinear phonon oscillator differs from the linear oscillator, which arouses our interests to research deeper in this field.

Optical Parametric Amplification of Femtosecond Laser Pulse Supervisor: ZHANG Jingdi / PHYS Student: YANG Yiguang / PHYS Course: UROP1000, Summer

This report includes two parts that I mainly pay attention to during this period. The first part is optical parametric amplification, including collinear OPA and noncollinear OPA, which can help to enhance the power of laser pulse. Also, a basic model for 3 stage OPA will be shown in the report. The second part is using MATLAB to do the simulation for photonic and ionic Raman scattering for diamond, ErFeO3 and BiFeO3. Raman scattering is a kind of inelastic scattering between photon and particle. During this process, energy exchange will happen, which result in the wavelength change of photon.

67

School of Science Department of Physics 68 Excitation and Imaging of Terahertz Phonon-polaritons Supervisor: ZHANG Jingdi / PHYS Student: CHENG Chun Hei / PHYS Course: UROP1100, Fall

The goal of this project is to study the effect of a sample placed on a crystal on the propagation of phonon-polariton in the crystal using FDTD (finite difference time domain) simulation which is helpful for the development of phonon-polariton spectroscopy. In this UROP 1100 programme, the dispersion relations of a thin (30μm) LiNbO3 layer in air with and without a sample on top are obtained from the 2D FDTD simulation of phonon-polaritons. This can potentially be used to study the effect of samples with different optical properties on the dispersion relations of the LiNbO3 layer.

68

Abstracts of UROP Projects 2021-22 School of Engineering

School of Engineering Department of Chemical and Biological Engineering

Department of Chemical and Biological Engineering Ultrasound-Triggered Drug/Gene Release from Self-Assembling Carriers Supervisor: CHAU Ying / CBE Student: SHIN June Yeol / BIOT Course: UROP1100, Fall

After observing that RNA was able to partition into Artificial Membraneless Organelles (AMO) successfully, whether low frequency ultrasound(LFUS) can induce the release of the RNA from the AMO was investigated. Exposure to LFUS for 1 minute actually increased the amount of RNA that is partitioned into the AMOs, rather than triggering the release from it. The effect of LFUS on cell viability and sonoporation was also conducted which showed that most cell lines are able to maintain cell viability for up to 5 minutes, while its cell membranes showed significant sonoporation. Such findings show prospects in utilizing LFUS for both the maximum loading of drugs into AMOs, and an efficient intracellular delivery of those AMOs through the LFUS-induced sonoporation of target cell membranes.

Development of Functional Food Supervisor: CHAU Ying / CBE Co-supervisor: LIU Yuanshuai / CBE Student: YORDANOV Matey Svetlan / CENG Course: UROP2100, Spring

One of the major concerns in all industries is that Earth resources are not endless and if not utilized wisely, we would not be able to preserve what we have now, for the future generations. Therefore, the focus is being changed from optimizing the industry purely for economic growth to optimizing it for more sustainable production processes.

This project focuses mainly on Food and Beverage Industry and development of upcycled products which do not only safe production waste, but also have health-benefiting properties. In this way the sustainability issue can be resolved by food waste valorisation.

As a result, multiple functional food and beverage products targeting this issue were developed.

70

School of Engineering Department of Chemical and Biological Engineering

Combinatorial Lipopeptides for Immunomodulation Supervisor: CHAU Ying / CBE Student: YAP Octavius Jessica / CENG Course: UROP1100, Spring UROP2100, Summer

Macrophages are widely circulated innate immune cells that play a crucial role in progression and resolution of inflammation. In response to its local microenvironment, macrophages can polarize into M1 phenotype which has pro-inflammatory characteristic, and M2 phenotype which perform anti-inflammatory feature. However, macrophage polarization disorder can damage tissue during inflammatory diseases and promote development of tumor. Hence, regulating macrophage polarization in specific disease could be a promising therapeutic strategy. Various peptide sequences are known to be responsible in macrophage polarization. These peptide signals were conjugated on the surface of micro-sized hydrogel and were presented as solid bound signals. The effect of peptide bound signal and its peptide combination on macrophage polarization were examined.

Combinatorial Lipopeptides for Immunomodulation Supervisor: CHAU Ying / CBE Student: LI Ho Wa / BIEN Course: UROP1100, Spring

Macrophages host the ability to generate different phenotypes via multiple pathways and acts as an important regulator in the development of cancer and different metastasis stage. M1 macrophages promotes inflammatory response and anti-microbial activities while M2 group macrophages serves as pro-healing and anti-inflammatory activator. (Jayasingam, S. et al., 2020; Orecchioni, M. et al., 2019; Gordon, S., 2017) Modulators including peptides on hydrogel can induce diverse polarization. However, the exact mechanism behind M1/M2 polarization and the effect of external factors including the morphology of the hydrogel is still unknown. To understand and predict this complex relationship for immunomodulation, hydrogel presenting peptides with different combinations are generated and different machine learning algorithms are built to understand its complex relationship between different properties of peptide, hydrogel and macrophage polarization.

71

School of Engineering Department of Chemical and Biological Engineering

Machine-Leaning Assisted Optimization of Pressure Swing Adsorption Processes Supervisor: GAO Hanyu / CBE Student: TAKARIANTO Abraham Aditya / CEEV Course: UROP1100, Spring

Pressure swing adsorption is an alternative technology to propylene separation that could reduce the capital and operational cost. A good adsorption process has a good selectivity towards the adsorbed product. In this project a system of propylene and propane is selected, and Z10-4 (Zeochem) is selected as the adsorbent for the model system. An isotherm of the propylene and propane is plotted. A numerical model of system operating at 650C with the pressure swing between 1 – 30 bar is plotted. The result shows that the pressure equation needs to be modified to remove the pressure lag and a better adsorbent or higher linear driving force needs to be selected to increase the efficiency of the adsorption process.

Active Learning for the Design of Polymerization Reactions Supervisor: GAO Hanyu / CBE Student: YANG Jianbo / CENG Course: UROP1100, Spring UROP2100, Summer

Polymer is an essential material widely used in our daily lives. Their properties can vary significantly due to their complex microstructures including composition ratio of monomers, length of polymer chain, position of sub-chains, the sequence of monomers and so on. The complicity opens a large area for researchers to design multifunctional polymers, also bringing a challenge task on how to relate parameters of microstructure to properties along with novel products. Machine learning, a cutting-edge data processing method, fits this area well because it enables researchers to process large-scale data in a shorter time and make predictions based on current examples. In the project, we continue the progress in last semester and try to optimize the execution time problem.

72

School of Engineering Department of Chemical and Biological Engineering

Generation of Adversarial Chemical Reactions Supervisor: GAO Hanyu / CBE Student: OH Jintaek / BIEN Course: UROP1000, Summer

Conventional generative models for the chemical reaction generation often uses the Variational Autoencoder (VAE). Whereas the Generative Adversarial Network (GAN) which is showing great performances in image and text generation, is not actively utilized as the generative model for the chemical reaction. This is due to the highly unstable training process and low reproducibility of GAN. Through the UROP 1000, different types of GAN were studied. Furthermore, latent-GAN which was originally intended for the novel drug design was examined and modified. Specifically, autoencoder part of the latent-GAN was reassembled to be used as the chemical reaction generation. In the future, reassembled autoencoder and GAN part will be trained and outputs will be analysed.

Designing of the Chemical Process for Polyethylene Furanoate-based Materials Supervisor: HU Xijun / CBE Student: YANG Jianbo / CENG Course: UROP3100, Fall

Now wood waste has accounted for a significant number of landfill capacity in Hong Kong, nearly up to 10%. Most disposed wood waste are wood pallets for cargo transportation and old-fashioned furniture. An applicable chemical process to transform wood waste into useful organic substrate is expected so that the pressure on landfill space in Hong Kong can be alleviated. Except for catalytic pyrolysis, there is also an advanced solution treating process that can fabricate transparent wood, which is a potential construction material. This process takes advantage of the hierarchical structure of wood and reduce influence on environment. The rest of the report will discuss about the transparent composite fabricated this semester and its potential modifications for more complicated functions.

73

School of Engineering Department of Chemical and Biological Engineering

Influence of Random Base in Poly(A) tail to the Function of Mammalian mRNA Supervisor: KUANG Becki Yi / CBE Student: DANI Raissa Gavrila / BIOT Course: UROP1100, Fall

mRNA or messenger RNA has a vital role in all organisms since it directly corresponds to the protein coding gene and allows protein to be produced within the cell. Post RNA transcription within the nucleus, the RNA will undergo a post-translational modification through splicing, 5’ 7-methylguanosine capping, and 3’ polyadenylation. A report by Chang et.al utilized the TAIL-Seq and they found that the poly(A) tail of mRNA has an insertion of other base pairs. Guanine insertion, in particular, was found favorable and can further stabilize the mRNA molecule. This report will thus further investigate the effect of inserting other base pairs such as cytosine residue in the poly(A) tail and observe its effect on the half life of the mRNA. RNAs with different poly(A) tails will be made through in vitro transcription and will be subjected for transfection into the mammalian cell. A luciferase assay will then be used to analyze the half-life of the protein by quantifying the luminescence of the oxyluciferin.

Influence of Random Base in Poly(A) tail to the Function of Mammalian mRNA Supervisor: KUANG Becki Yi / CBE Student: SHI Naiyu / BIEN Course: UROP3100, Fall

The COVID-19 pandemic has brought messenger ribonucleic acid (mRNA) vaccines to the general public. mRNAis a single-stranded RNA. It delivers the genetic information of the DNA to the cytoplasm to synthesize proteins. To protect mRNA from degradation, a poly(A) tail is added at the 3’ end. Despite the name suggesting it is a string of adenines, random insertions of guanine (G), cytosine (C) and uracil (U) inside poly(A) tails have been found. These insertions seem to have the capability of affecting the half-life and the translation efficiency of the mRNA, however, whether these insertions are the reason for the changes and how they influence the functions and properties of mRNA are not known. This project aims to investigate the influence of insertions of G in poly(A) tails of mammalian mRNA.

74

School of Engineering Department of Chemical and Biological Engineering

The Construction of Dual-functional Synthetic RNA Control Unit Supervisor: KUANG Becki Yi / CBE Student: WONG Hiu Ching / CBGBM Course: UROP1100, Fall

Techniques of polymerase chain reaction (PCR), gel electrophoresis, urea-PAGE, in vitro transcription (IVT), cell passage and transfection are explored and investigated in this UROP project. By performing a series of experiments, it is shown that by using these techniques, any DNA fragments containing desired genes can be amplified, fused and transcribed before being introduced into cells for further expression and growth given the correct conditions and procedure. Through this UROP project, these wet-lab skills have been acquired and practiced, and can be further applied to future researches in fields such as genetic function and cell type characterization, as well as protein synthesis and drug discovery. characterisation

The Construction of Dual-functional Synthetic RNA Control Unit Supervisor: KUANG Becki Yi / CBE Student: YAM Tsz Yui / BIEN Course: UROP1000, Summer

Over the past few decades, deoxyribonucleic acid (DNA) and ribonucleic acid (RNA) has gone from an enigmatic molecule with presumed structure and functions within the nucleus to one of the hallmarks of modern biology. There are now many studies regarding either DNA or RNA nowadays, for instance, DNA insertion and RNA modification. The aim of this project is to master the skill to expand plasmids in bacterial cells. Besides, the second aim is to master the skill to synthesize DNA templates and obtain corresponding message RNAs (mRNAs) from in vitro transcription (IVT) with high purity.

Improving Data Analysis Methods for Shotgun Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: CHAN Tsz Fung / BIEN Course: UROP1000, Summer

Spectral library search is one of the effective and efficient approaches for peptide identification. With the aid of the developed machine learning models, massive tandem mass spectra can be predicted and construct a spectral library for matching with the sample spectrum to identify the peptide sequence by their similarities. The purpose of this study is to validate the spectra predicting models and analyze the implication of machine learning on spectral library search for improving the effectiveness of peptide identification. This study will apply various spectra-predicting machine learning models in building spectra libraries and evaluate the performance of the spectral library built with the predicted spectra with the real experimental data.

75

School of Engineering Department of Chemical and Biological Engineering

Improving Data Analysis Methods for Shotgun Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: CHUNG Chun Kit / BIEN Course: UROP1000, Summer

The rapid development in proteomics promotes a more complete study of the roles of proteins in living organisms. With the limitations encountered by database search and spectral library, spectral archive has emerged as an improved tool to handle and utilize the tremendous amount of data produced from mass spectrometry (MS). This report compares two common deep learning based tandem mass (MS/MS) spectrum prediction tools, Predfull and Prosit, by analyzing their underlying network architectures and performing real searching on the datasets obtained from two species. The comparison shows that Predfull outperforms Prosit with an average 1.42% increase in accuracy. Two reasons are proposed to explain this result.

Improving Data Analysis Methods for Shotgun Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: LI Ka Yan / CEEV Course: UROP1000, Summer

Tandem mass spectrometry (MS2) is a common method used in proteomics. The identification of peptide sequences in biological samples creates opportunities for advancement in the medical industry, such as early diagnosis of diseases and development of new drugs. Protein sequencing can be achieved by peptidespectrum matches (PSMs), by using automated algorithms to analysis the mass spectra. There are currently three major approaches to identify the mass spectra – database searching, spectral libraries and spectral archives. This project will focus on searching by spectral archives, which is a relatively new approach that is still under development. Making use of Facebook AI Similarity Search (FAISS), 3 basic indexing methods are being selected in constructing spectral archives in this project, to investigate the effects on the time and accuracy of identifications by spectral archives built with: 1) different numbers of indices and 2) different indexing methods.

76

School of Engineering Department of Chemical and Biological Engineering

Improving Data Analysis Methods for Shotgun Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: SHAO Heyan / CENG Course: UROP1100, Summer

Tandem mass spectrometry has emerged as a state-of-the-art method to detect and measure protein samples in the form of digested peptide sequences. Determining the data quality of peptide fragmentation is one of the major challenges in proteomics, as the peptide fragmentation behavior may differ in practice. This research presents metrics for measuring the quality of peptide fragmentation based on tandem mass spectra characteristics and investigate the peptide fragmentation quality based on the peptide sequence only. By examining the peptide fragmentation performance, we can assess the importance of peptide features, which helps to improve the quality of peptide identification. It is investigated that K-ending peptides showed better fragmentation pattern compared to R-ending peptides, and both polar and aromatic amino acids have no obvious correlation with the quality of fragmentation.

A Bifunctional Reactor Design for Biological and Chemical Oxidation Processes of The Greywater Remediation

Supervisor: LAM Leung Yuk Frank / CBE Student: MA Man Chun / CENG Course: UROP1000, Summer

Water scarcity make water recycling become an important task around the world. Greywater treatment is one of the water recycling methods. It is essential to achieve sustainable development and enhance the efficiency of the water recycling. This project also tries to reduce the operating cost of the greywater treatment machine. A total of six step will be adopted in this greywater treatment machine, including Foam, two ozonation, one biotreatment, one sedimentation and one UV disinfection. The TOC value is mainly used for the measure the purity of the water reclaimed. This paper will provide part of the treatment result and some parts of the treatment modification.

77

School of Engineering Department of Chemical and Biological Engineering

Design and Development of Functional Food Supervisor: LIU Yuanshuai / CBE Student: KAO Tzu Yuan / BIEN Course: UROP1000, Summer

The experiment applies food processing technology to enzyme tiger nuts and extracts its liquid by (1) blending tiger nuts powder with water to get smooth slurry; (2) Hydrolyzing the slurry with α-amylase and glucoamylase; (3) adjust acidity by adding potassium phosphate. The αamylases and glucoamylase which are used to saccharify tiger nuts can achieve a high enzymatic hydrolysis result, including high reducing sugar amount, residues, and high degree of protein hydrolysis. Tiger nuts contain high calcium which help prevent cardiovascular diseases, diabetes, cataracts, bone deformation, and tissue damage caused by environmental hazards, as well as regulate the production and release of insulin. The tiger nut liquid can be directly consumed, and it can replace milk to make dairy products for those who have lactose intolerance and vegan people.

Design and Development of Functional Food Supervisor: LIU Yuanshuai / CBE Student: WAN Ching Natalie / BIEN Course: UROP1000, Summer

This report is about the preparation of vegan gelato which involves the aspect of food processing. The base of the gelato is made from enzymatically treated tiger nut and oat powder. Ingredients such as soymilk powder and coconut oil are then added to the base of the gelato and is then heated until smooth without any clumps. Erythritol is used as a substitute for white sugar to provide sweetness for the gelato. Liquid nitrogen is then added to the base to form gelato. The high amount of vitamins, minerals and amino acids in the gelato also has a lot of health benefits and is suitable for all people, even vegans and those with lactose intolerance.

78

School of Engineering Department of Chemical and Biological Engineering

Humanitarian Chemical Engineering - A Summer Expedition Supervisor: MAK Andrew T C / CBE Student: HUI Yanna Yatnga / BIEN Course: UROP1000, Summer

Malaria, a fatal disease caused by parasites, has put half of the world’s population at risk. Together with the continuous ENGG 4930E SIGHT project of Myanmar Malaria, this UROP 1000 project focuses on humanitarian engineering solutions on Malaria Control, not only Myanmar but globally available approaches. Knowing that the focus of humanitarian engineering is to serve a highly effective solution at a low cost to suit everyone in need, a total of three approaches as options at different stages during design thinking have been proposed. Including the Water Recycle Robot, Water Source Improvements, and Bacteriophage approaches. The purpose of these approaches are all to eliminate Malaria at root for the highest effectiveness and alleviate its harm to human beings. These solutions are compared and analysed under cost, possibility of implementation, and effectiveness, with a final decision of focusing on developing the idea of Water Recycle Robot.

Neural Circuits for Visual Behaviors in Zebrafish Supervisor: SEMMELHACK Julia Lee / CBE Student: ZHAO Yicheng / BCB Course: UROP1100, Summer

Behavior studies are crucial for research about the brain, here we used zebrafish as a model to study an important part of its hunting behavior, strike. Prey captures of a zebrafish consist of 3 steps, J-turns, approaches, and capture strikes. Strike is the last step of prey capture, and usually appears when a strong prey capture occurs (typically characterized by strong J turns, i.e. the bend of the larvae tail). To simulate the prey, we used an LCD screen to present the larvae with a moving dot. We tested 3 different types of dots, sweeping dots, stationary dots, and flickering dots respectively; the size of which was also tested to find the ideal condition for triggering prey capture and strike. We found that a 4-degree dot is optimal to trigger both prey capture and strike, and a sweeping dot is more effective in inducing behaviors than stationary and flickering dots.

Figure 1: Strike of an embedded fish vs a free swimming fish

79

School of Engineering Department of Chemical and Biological Engineering

Neural Circuits for Visual Behaviors Supervisor: SEMMELHACK Julia Lee / CBE Student: FARHAN Ishraq / IRE Course: UROP1000, Summer

Freezing behaviour, expressed by bradycardia, can be provoked by visual stimulus. The project seeks to establish that this behaviour can be exhibited and to optimise the features of a visual stimulus required to evoke this response. The stimulus evoking freezing is a moving dot.

behavior optimize

One feature of the stimuli to study is the changing luminance as it moves. With a dynamic luminance stimulus, which is a rectangle spanning the screen whose contrast is changing over time, we control for the changing luminance received by the eyes to determine whether the freezing behaviour can be ascribed to the changing luminance of a dot as it moves across the larvae’s field of view.

Anode Materials for Lithium Ion Batteries Supervisor: SHAO Minhua / CBE Student: CHOI Jae Uk / CENG Course: UROP1100, Fall UROP2100, Summer

Higher energy density and higher voltage batteries have been the long sought out characteristics for further development of electric vehicles (EV) and electrical storage systems (ESS). However, current Ni-rich LiNi0.8Co0.1Mn0.1 (NCM811) cathodes exhibit oxidative instability under high potential (>4.2V) conditions, resulting in particle surface deterioration and poor cycling stability. In this report, Li-rich Li1.2Mn0.54Ni0.13Co0.13 (LR-MNC) cathode and bis(2, 2, 2-Trifluoroethyl) Carbonate (DTFEC) electrolyte were explored. LR-MNC exhibits promising stability and capacity at higher potential (<4.8V) with the higher Mn content. Highly fluorinated carbonate DTFEC offers to be a good replacement for ether-based electrolyte such as 1,2- dimethoxyethane (DME) and 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropylether (HFE), which is unstable when paired with high Ni-content cathodes at higher potential.

80

School of Engineering Department of Chemical and Biological Engineering

Anode Materials for Lithium Ion Batteries Supervisor: SHAO Minhua / CBE Student: JUNG Euntaek / CENG Course: UROP1100, Fall

Lithium metal is considered as a novel anode material due to its high specific capacity and low voltage plateau. To enable practical application of Li metal anode, problems regarding volume expansion and side reaction between active lithium and electrolyte need to be solved. Adjusting lithium salts and solvents of electrolyte is one way to address the limitations of Li metal anode. In this project, a ether based electrolyte containing lithium difluoro(oxalate)borate (LiDFOB) salt, 1,2-dimethoxyethane (DME) and 1,1,2,2-tetrafluoroethyl-2,2,3,3-tetrafluoropropylether(HFE) solvents was prepared. Cells were assembled with LiCoO2 cathode, Li meatal anode and such electrolyte. During the electrochemical measurement, specific capacity and coulombic efficiency of the cells were measured for 100 cycles. Such electrolyte was stable on both LiCoO2 and Lithium metal sides, leading to high coulombic efficiency (99.5%) and capacity retention ratio (97% after 100 cycles).

Anode Materials for Lithium Ion Batteries Supervisor: SHAO Minhua / CBE Student: KIM Sungrae / CHEM Course: UROP1000, Summer

Currently, aluminum foil is widely employed as a cathode current collector due to its high electronic conductivity, low cost, and chemical stability. Despite its notable characteristics, the aluminum current collector still has a few obstacles to overcome to meet the demands of next-generation lithium batteries: higher chemical stability in high voltages and enhanced electronic conductivity. The study employs a carboncoated aluminum foil as a current collector in the lithium metal battery with Li1.2 Mn0.54Ni0.13Co0.13O2 (Li-rich) cathode. Half-cell tests with lithium metal batteries are performed to observe the enhanced capacity delivery and cycling stability due to the employment of carbon-coated aluminum as a new current collector material.

81

School of Engineering Department of Chemical and Biological Engineering

Anode Materials for Lithium Ion Batteries Supervisor: SHAO Minhua / CBE Student: KWAN Yee Tung / SUSEE Course: UROP1100, Spring

Due to the rising demand for higher energy density batteries, interest arises in the implementation of highvoltage cathodes in the field of Li-ion batteries, such as LiNi0.5Mn1.5O4 and Ni-rich layered oxides. However, the electrochemical stability windows of prevailing electrolytes are insufficient to withstand such a high voltage operation. Highly concentrated electrolytes have been a notable strategy to deal with the stability issue due to the special structure and improvement in electrolyte stability. In this project, Lithium batteries with the LiNi0.8Co0.1Mn0.1O2 cathode are assembled with 1,3-dioxolane as a solvent and LiTFSI as a salt with different salt to solvent weight ratios. Cycling tests of 100 charging and discharging cycles are performed on the batteries to evaluate their capacity retentions.

Anode Materials for Lithium Ion Batteries Supervisor: SHAO Minhua / CBE Student: SO Kwok To Timothy / SENG Course: UROP1000, Summer

To improve the next generation of Li-ion battery, it is very important to maintain the stability and efficiency of the battery. NCM811 battery has been a common choice due to its high energy density and being cost effective, however it suffers from poor cycling stability. There are many projects on how different electrolytes could improve this. Particularly lithium hexafluorophosphate (LiPF6) and lithium bis(trifluoromethanesulfonyl)imide (LiTFSI). However, these electrolytes, particularly LiTFSI causes corrosion on the aluminium current collector which decreases its efficiency and stability. In this project, the effect of carbon coated aluminium will be explored to see whether it does improve the corrosion issue present in the current collector. The cells with NCM811 cathode, Li metal anode would be assembled and subjected to charging and discharging cycle 75 times to test the effect of carbon coating. Aluminum

82

School of Engineering Department of Chemical and Biological Engineering

Advanced Catalysts for Water Splitting and Hydrogen Production Supervisor: SHAO Minhua / CBE Student: CHAN Yik Yu / SUSEE Course: UROP1100, Spring

Hydrogen is a promising and clean energy source, and it is a possible alternative to fossil fuel. However, the development of active and cheap electrocatalysts to replace Platinum remains a challenge. Among the alternative metals, ruthenium stands out because of its prominent catalytic activity, pH universal application, and cheaper cost compared to the precious metal family. Herein, ruthenium-based catalysts supported on a carbon matrix doped with different nonmetallic elements, including sulphur (S), nitrogen (N) and boron (B) were studied. Some general background will be given in this paper. Subsequently, the method of this study will be provided. Finally, the result will be discussed.

Organic Molecules for Hydrogen Storage Supervisor: SHAO Minhua / CBE Student: CHAN Shun Hong / SUSE Course: UROP1100, Summer

In this report, the importance of decarbonizing the energy system is addressed. Several solutions for energy carriers are mentioned. Of all of them, hydrogen was chosen to be studied. A further emphasis was put on liquid organic hydrogen carriers (LOHC in short) as it is the most efficient in terms of gravimetric density and volumetric density and has less flaws. By taking references to current successful LOHC, few key characteristics are generalized for the criteria of a good LOHC. Moreover, a structure which consists of N-heterocycles and a 5-ring structure are found to have lower dehydrogenation temperature. A few potential carriers are synthesized with the above characteristics. Further work is required to confirm its potential as a LOHC.

Reusable Protein-Based Materials for Uranium Extraction from the Ocean Supervisor: SUN Fei / CBE Student: LEE Juwon / CBME Course: UROP1100, Spring

The paper ‘B12-induced reassembly of split photoreceptor protein enables photoresponsive hydrogels with tunable mechanics’ utilizes the photoresponsive characteristics of newly designed proteins to address the limitations of currently existing proteins.

83

School of Engineering Department of Chemical and Biological Engineering

Analyzing the Possible Water-Soluble Proteins for Deep-Tissue and Cancer-targeting in Vivo Imaging Using a Photoacoustic Imaging System Supervisor: WONG Tsz Wai / CBE Student: LAI Ka Wai / BIEN Course: UROP2100, Summer

The objective of this project is to develop a biodegradable contrast agent for a deep-tissue photoacoustic imaging system which can potentially overcome various limitations in other existing imaging modalities and generate images with higher qualities. This system requires a contrast agent that is biodegradable, non-toxic, has a near-infra-red absorption peak, a high extinction coefficient, and a low quantum yield. A potential candidate for such a contrast agent is composed of water-soluble chlorophyll-binding proteins, SpyTag and SpyCatcher proteins, and human epidermal receptor 2 affibodies. Traditionally, chlorophylls are used as the central phytochrome binding to the protein. However, in this experiment, LvPCPS was combined with bacteriochlorophyll A for the formation of the contrast agent. It was found that bacteriochlorophyll is a suitable alternative to chlorophyll from plants for the contrast agent.

Mask but Not the Least Supervisor: YEUNG King Lun / CBE Student: CHEUNG Howard Yau Ming / CIEV KANDWAL Inika / EVMT LIEM Albert / SUSEE Course: UROP1100, Fall UROP1100, Fall UROP3100, Fall

The COVID-19 pandemic has not only affected and hindered our daily lives, or the environment today, but has asked us to seriously consider the environmental impacts of our lifestyle choices. A prominent issue at the moment is the unattended disposal of face masks, around 129 billion face masks every month. This paper explores the creation of composite prototypes using shredded mask waste and other waste materials (paper, natural fibres etc) to create load-bearing, water resistant and aesthetic material. We aim to create products of high-value and function as well as using them to raise awareness around the importance of proper mask disposal.

84

School of Engineering Department of Civil and Environmental Engineering

Department of Civil and Environmental Engineering Experimental Testing and Material Characterization of Bamboo Structural Members Supervisor: DIMITRAKOPOULOS Ilias / CIVL Student: TING Chung Lam / CIVL YAN Tak Po Jonathan / CIVL YIM Kwan To / CIEV Course: UROP1100, Fall UROP1100, Fall UROP1100, Fall

Bamboo is a sustainable and green building material. It is environmentally friendly as its production process and growth produces little greenhouse gases. Its quick-growing behaviour, compared to other plant-based materials like wood, provides high sustainability. In view of convenience, bamboo is abundant and easy growing, requiring less transportation of building materials, time and money costs. However, there are concerns about bamboo’s strength and consistency, making usage of bamboo as a material for permanent structure rare.

The project aims to develop a calculation of shear and tensile strength of bamboo based on its physical properties, like diameter, thickness and moisture content. This project focuses on two different specimens: Bomnhusa blumeana, and Dendrocalamus, asper, and perform shear test, tension test, and split-pin test to obtain data for evaluation. The data are used to determine the relationship between different physical characteristics of bamboo and its strength.

The shear, tensile strength and split-pin tests are done according to the ISO22157-2019(E) standard. Orientations of bamboos, top, middle and bottom parts, are marked, as well as other properties, thickness and diameter. Diameters are taken along 2 perpendicular directions and the length of specimens is measured at 4 spots. The average values are used in calculation so that human errors can be eliminated.

To conclude, the results in the report shows that bamboo is a brittle material and has low resistance to bending moment. It is too unsafe to use bamboo as a permanent building material at this stage. More research is needed before putting bamboo in use for construction.

85

School of Engineering Department of Civil and Environmental Engineering

Analysis of Compound Extreme Using Cmip6 Multi-Model Ensemble Supervisor: IM Eun Soon / CIVL Student: ZHOU Zixuan / EVMT Course: UROP1100, Spring

The demand for accurate climate projections has grown significantly because people realized the importance of projecting climate change impacts for societal adaptation. However, climate models have systematic bias, hindering people from directly using them for impact analysis. Bias correction is widely used to produce unbiased climate projections, including component-wise and direct bias correction. This project focuses on bias correction for heat-stress indicators in South Korea using multiple regional climate model simulations. We will compare five bias correction methods in this study, including four univariate (direct correction) and one multivariate (component-wise bias correction) approach, for correcting two heat-stress indices, Wetbulb Globe Temperature (WBGT) and Apparent Temperature (AT). These two indices have different intervariable dependencies between temperature and relative humidity. Specifically, I was responsible for producing the Quantile Delta Mapping (QDM) for Wet Bulb Temperature. I went through two stages. Firstly, I remapped the historical simulation (1979-2014) and future simulation (2015-2100) based on time interpolation to resolve non-standard calendar and spatial interpolation to the ERA5 grid. After that, I applied QDM to future simulations based on the error calculated between ERA5 and historical simulation.

Characterizing Driving Behaviors of Human Drivers When Following Automated Vehicles Using The Realworld Dataset Supervisor: JIAN Sisi / CIVL Student: CHANG Sin Tong / CIVL Course: UROP1100, Spring

It is apparent that the usage of automated vehicles on road are increasing from the trend of market growth. Before the replacement of human-driven vehicles (HVs) by automated vehicles (AVs), there exists a transition period where there are mixed traffic streams composing of both AVs and HVs. This presents a research opportunity as the interactions between AVs and HVs, primarily the changes in behavior of HVs when following AVs have not been fully established and understood. This is particularly important as comprehensive understanding will be required to ensure the design of AVs driving algorithms are adequate in creating a safe driving environment for road users. A primary goal is to increase safety by the introduction of AVs on the road.

86

School of Engineering Department of Civil and Environmental Engineering

Recyclable Hydrogel-Based Brick for Construction on Mars: A Feasibility Study Supervisor: QIU Jishen / CIVL Student: NGAN Man Cheuk / CIGBM Course: UROP4100, Fall

This study aims at investigating the efficiency of different recycling approaches for the hydrogel-based bricks and their impact to the gel’s mechanical properties after recycling. Our research object is Gelatin as it provides high recyclability with little requirement of additional chemical agents and environment condition. Compression tests and SEM analysis are conducted to examine the impact of recycling bricks to the stress, elastic modulus, and failure modes of the bricks.

The 2 nd Generation bricks made with the recycled materials from the 1st generation bricks and water exhibit similar mechanical properties and can achieve a compressive strength of 9.083MPa, which is about 87.8% of the original strength before recycling. Another recycling method in which the hydrogel is extracted into a solution form and mixed with new ceramic beads can maintain a compressive strength of 6.55MPa, which is about 63% of the strength of 1st generation bricks.

A Cement-Free Novel Concrete that Absorbs Greenhouse Gas Co2 to Heal Itself and Improve its Mechanical Performance. Supervisor: QIU Jishen / CIVL Student: LEE Yin Ni / CIGBM Course: UROP1100, Spring UROP2100, Summer

Reactive magnesia (MgO) is a promising alternative to conventional Portland cement (PC) as it could potentially recapture carbon emissions from concrete production. However, inadequate carbon dioxide (CO2) diffusion for carbonation and thus strength development remains a major technical challenge for the industrial application of reactive MgO cement (RMC) as concrete binder. Previous works have found hollow natural fibres (HNFs) effective in establishing CO2 transportation channels to enhance CO2 penetration and thus carbonation in RMC-based composites, however, such channels might be blocked by water, especially in composites with high water-to-binder (w/b) ratios. To counteract this effect, RMC-based composites were dried for different durations after demoulding to lower free water content and facilitate carbonation at core regions in the subsequent curing process. This study found that a balance in w/b ratio and drying duration could provide the ideal formulation of HNF-added, RMC-based composites with enhanced carbonation, such that RMC-based composites could potentially substitute PC-based matrix as a net-zero concrete with the possibility of complete utilisation of RMC’s carbon sequestration capability.

87

School of Engineering Department of Civil and Environmental Engineering

Human-Robot-AI Symbiotic Mobile Mapping Solution for Fast and Regular Examination of Civil and Environmental Engineering Infrastructures Supervisor: WANG Yu-Hsing / CIVL Student: CHAN Kin Yan / CIVL Course: UROP1100, Spring

Artificial Intelligence, machine learning, deep learning are trendy techniques widely expending to various fields and providing deeper insight through development. This project aims for exploration of opportunity of the application and implementation of the well-developed machine learning techniques on civil engineering fields, including maintenance of concrete structure and other topics. For exploration and knowledge development purpose, full model development process from data collection, data processing, model training, prediction performing, model fine-tunning and model performance assessment was studied. Then various possible machine learning library including sklearn, tensorflow Keras for both supervised learning and unsupervised learning requirement was explored in detail for opportunity of merging with domain knowledge. U-net architecture used for semantic image classification was also investigated and prepared for future projects including classification of transport types.

Human-Robot-AI Symbiotic Mobile Mapping Solution for Fast and Regular Examination of Civil and Environmental Engineering Infrastructures Supervisor: WANG Yu-Hsing / CIVL Student: LI Sau Long / CIVL Course: UROP1100, Spring

In Hong Kong, more than 30,000 old buildings and infrastructure demand routine checking and maintenance, some of which require urgent repairs. Manual visual inspection remains to be the standard procedure in the civil engineering and related industries, which could be prone to errors, time-consuming and labour-intensive. Compounded by insufficient manpower, economic losses and safety issues, current practices might prove to be inadequate to meet the city’s operational demands. In this project, a Convolutional Neural Network architecture is utilised to analyse and categorise concrete defects into 3 different levels of damage, in order to speed up the inspection process. A state-of-the-art VGG16 model is incorporated into the code which has approximately 138 million parameters which gives significant flexibility to fine-tune. The results of the accuracy of the model is summarised into a confusion matrix and IoU (Intersection over Union) scores to assess the performance of the model on future projects and identify potential errors sources.

88

School of Engineering Department of Computer Science and Engineering

Department of Computer Science and Engineering Video Analytics and IoT People/Asset Sensing for Smart City Applications Supervisor: CHAN Gary Shueng Han / CSE Student: CHANG Hong-yuan / DSCT Course: UROP1100, Fall

Current large factor super-resolution (SR) models are degradationsensitive, that is, when the degradation for testing images differs from the degradation for generating training LR-HR pairs, the model will perform poorly. This largely limits the practical usage of the SR model since, in general, it is hard to know the exact degradation process of a low-resolution (LR) image in a real-world setting. To alleviate this issue, we propose a frequency-based domain adaptation module that maps images from different degradation domains to the same domain by applying a learned mask on the image in the frequency domain. The module can then be trained on top of other SR models to improve the model’s performance when facing degradation domain mismatching.

Video Analytics and IoT People/Asset Sensing for Smart City Applications Supervisor: CHAN Gary Shueng Han / CSE Student: TSANG Hong Ting / COMP Course: UROP1100, Summer

The objective of this project is to create a standard for Pervasive Positioning, allowing applications to locate an object anywhere, no matter indoor or outdoor, on a country scale seamlessly. In current technology for location-based applications, both indoor and outdoor positioning work well independently but there is no one standard that allows applications to perform Pervasive Positioning using current technology. Utilizing the approach of decoupling and then recoupling, the lookup server first gathers all data and server addresses of site owners, depending on the privacy control and requested data format, set by site owners and applications, returning the appropriate data or address for further computation. Under this approach, we can achieve Pervasive Positioning utilizing current technology by contacting the right party using the right formats. In this semester, the API (Application Programming Interface) for SDK (software development toolkit) is in development. Using the scheme provided by Mercator projection, Google Map Grid Reference System, and Multi-Layered Space Model, API can respond and return data from the database to SDK.

89

School of Engineering Department of Computer Science and Engineering

Video Analytics and IoT People/Asset Sensing for Smart City Applications Supervisor: CHAN Gary Shueng Han / CSE Student: ZHOU Yanting / MATH-CS Course: UROP2100, Spring

Pervasive positioning is to locate an object anywhere seamlessly on a country scale. With current positioning technologies’ capability to support both indoor and outdoor environments, Pervasive Positioning Standard aims to bridge every party together. This standard specifies a set of communication protocols, data organization and format definitions for site signals, and data organization for maps using existing map standards. With this standard, existing location-based service (LBS) applications are able to operate anywhere in Hong Kong including their original supporting zone. In this report, I will briefly introduce the predesigned framework of the whole standard, and then elaborate more on the SDK APIs for applications, which is the part I worked on.

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: CHANG Joshua / QFIN CHAU Yu Foon Darin / MATH-IRE Course: UROP1000, Summer UROP1100, Summer

In this UROP project, we worked on video crowd counting using deep learning1 . We reorganised and restructured the code from a novel video crowd counting approach, PHNet, and trained the network using 2 video crowd counting benchmarks, Venice and Crowdflow. We then made modifications to the network, added some more data augmentation options, and also developed our own synthetic video crowd counting dataset using Unity and trained it with the network. organized

90

School of Engineering Department of Computer Science and Engineering

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: CHEN I Chieh / COMP Course: UROP1100, Spring

Crowd counting has been an important technique in urban planning and public safety, and it has been a popular field in artificial intelligence. Initial works had tried to solve the problem with detection-based object counting approaches; however, these methods would suffer from data with dense crowd or objection occlusion. Recent works have been proposing regression-based methods and have achieved success and overcome the hazard above. In this project, we are going extend this problem to handle nighttime crowd data with existing crowd counting models and discuss possible way to improve the models’ performance on nighttime data. We will focus this problem particular at the domain of data processing and further learning or modification on existing models.

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: DAI Tianyuan / COSC Course: UROP2100, Fall

Crowd counting is to estimate the number of target objects (e.g., person, cars) in crowd-scene images or frames of videos. Current state-of-art methods depend heavily on supervised learning but the data is scarce. It adopts the assumption that all data should be identically and independently distributed (a.k.a., i.i.d.). However, data in the world often violate this i.i.d. assumtion, having different scales and distributions. This induces super difficulty in transferring crowd counting models trained on a particular dataset to other datasets. Therefore, how to do Out-of-Distribution (OoD) generalization in crowd counting to boost the performance of crowd counting models across different datasets becomes a great demand. In this paper, two main methods are investigated to do Out-of-Distribution generalization in crowd counting tasks: 1) through semi-supervised/unsupervised learning; 2) through dataset construction. An experiment is conducted to show that our dataset outperform current datasets, which shows that it could be adopted to boost the performance of crowd counting models across different scenes and datasets.

91

School of Engineering Department of Computer Science and Engineering

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: HO Ka Shuen / COMP Course: UROP1000, Summer

Global Navigation Satellite System (GNSS) provides positioning service that determines users’ location using data received from satellites. With a more accurate system, new user experience could be provided and urge for development of innovation especially in the era of intelligence technology and automation. This report explores ways to analysis smartphone location data and produce higher accuracy positions with decimeter or even centimeter resolution. Two approaches were examined, weighted least square line with outlier detection and coordinate with nearest neighbors. Location measurement error was reduced as a result but there is more algorithm to be explored that could minimize error more significantly.

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: KAO Shiu-hong / DSCT Course: UROP1100, Fall

The traditional architecture of the Convolution Neural Network (CNN) has been one of the most common and powerful neural networks, especially for dealing with image information. However, some of its shortcomings, such as redundant filters and parameters, are believed to exist, contributing to unnecessary computation, and reducing the efficiency of the network. Dynamic neural networks, on the other hand, provide a more flexible architecture based on different samples.

In this report, traditional Convolution as well as the other two pixel-wise dynamic networks, specifically, Involution and Conditionally Parameterized Convolutions (CondConv), are introduced. The basic architectures of these networks are firstly specified. Afterward, an experiment will be conducted to compare the efficiency and performance of distinct networks.

92

School of Engineering Department of Computer Science and Engineering

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: LI Ho Yin / CS Course: UROP1100, Fall

Since generating reference points in floor plans is essential in indoor localization, this report analyzes approaches in floor plan recognition that can help to generate reference points by computers automatically. In particular, this report mainly investigates a deep learning network which is proposed by a previous research team and tries to perform preprocessing on the input image to help the pre-trained model generate a better prediction on the floor plan in HKUST. Although the prediction result is still far from successful to generate RPs, the features of the model will be discussed, and some insights will be provided which hope can help to the further development of indoor localization.

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: LIU Ming Hay / MAEC Course: UROP1000, Summer

Pervasive positioning aims to locate users seamlessly, whether indoors or outdoors. Currently, different formats of data and applications are needed for users to locate themselves, while users are entering another building or going outdoor. The implementation of pervasive positioning requires the correct format and correct parties to request location services. The Pervasive Positioning Standard states the data organization and defined formats for site signals and maps. A database would store the data provided by the site owners in the specified format. With this standard and the help of the site owners, users can have a more convenient experience with location applications.

93

School of Engineering Department of Computer Science and Engineering

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: PENG Zhuoxuan / COMP Course: UROP4100, Fall UROP4100, Spring

Crowd motion analysis is crucial for real life applications such as crowd surveillance, but current mainstream tasks like crowd counting ignore the dynamic features and provide limited information for reference. Therefore, in this project I propose flow maps, a brand new data format combining density maps and optical flows, to monitor crowd motion while not suffering from problems occurred when directly applying optical flows on crowd data. I also construct a new model, FlowCounTR for flow map estimation. Experiments show that the performance still need improvements, but it will be a potential direction in the area of crowd motion analysis.

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: XIANG Letian / COMP Course: UROP4100, Fall

Dementia is a set of symptoms including loss of memory, and impaired decision-making abilities. It may cause severe physical, psychological, social, and economic impact, and affect not only the patient, but also their families, caregivers, and even society. To support people in need, we apply mobility analytics on a large group of patients using location data. Given the patients’ trajectories, we try to understand their activities and achieve the aim of anti-wandering to improve the quality of life of the affected people. We would consider the spatial and temporal aspects, and we would also conduct lost analysis on the records of wandering patients.

94

School of Engineering Department of Computer Science and Engineering

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: XIA Zhiqiu / DSCT Course: UROP1100, Spring UROP2100, Summer

This semester, I enrolled in UROP 2100 to continue my work on the project AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data, supervised by Prof. Chan. In particular, I read the code of RoNIN carefully and reproduced its result smoothly. Based on that, Mr. He and me are building a system about personalized Pedestrian Dead Reckoning (PDR). This report mainly introduces the project background, project challenges, key observations, and several experiments with results. In the end, this report will briefly introduce our ideas and expectation for future work, which is left for UROP 3100.

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: YAO Chongchong / DSCT Course: UROP1100, Spring

2D point-formed data models a lot of real problems and account for a large portion of realworld geographical data, on which a common and natural requirement is to perform range/region queries. To meet such requirements, an efficient algorithm for range querying should be proposed. R-tree is known as an indexing mechanism that can efficiently search non-zero size multi-dimensional data. In this project, I study the structure of R-tree, implement a modified version of R-tree that operates on 2D point-formed data, and test and visualize its performance on both randomly generated and specialized real-world data. Eventually, I come to the conclusion that R-tree is a powerful tool for handling range queries on point data when the search range is reasonable.

95

School of Engineering Department of Computer Science and Engineering

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: YANG Lin / COMP Course: UROP1100, Spring UROP2100, Summer

Image restoration is an elementary tool for numerous low-level visionrelated tasks. It involves generating a high-quality clean image from a given low-quality degraded version. The degradation may occur during the picture capturing, transmitting, and preserving, resulting in different restoration tasks, such as denoising, deraining, dehazing, and deblurring. In this article, the development of image restoration tasks is reviewed, and two representative models will be focused on. One of them is an MLP-based model, which shows stateof- the-art performance on various restoration tasks, and the other one is a transformer-based model, which is all-in-one and degradation-blind. In the end, the future challenges of image restoration will be discussed.

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: ZHOU Yanting / MATH-CS Course: UROP1100, Fall UROP3100, Summer

Pervasive positioning is to locate an object anywhere seamlessly on a country scale. With the aim of connecting every party involved in a localization process, this standard contains a set of communication protocols, data organization and specification for site signals, and map data organization standard. In this summer semester, we carried on the development of the SDK APIs for application developers and corresponding Web APIs for site owners to respond queries. More specifically, we completed the preliminary development of the SDK APIs for initialized handshaking, indoor localization, and map display. The SDK functions give reasonable outputs when we tested them under the pre-set site servers and real collected data.

96

School of Engineering Department of Computer Science and Engineering

AI Meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: ZHU Boan / CPEG Course: UROP1100, Summer

Fisheye images are often distorted and twisted, thus they often require distortion rectification. The generation-based method is one of the most popular methods because of its label-free training and onestage rectification. However, it has two main problems. The first one is that the decoder of the network is overburdened for rebuilding the structure and content of the images simultaneously. The second problem is that the naive skip-connection delivers the images directly which may lead to distortion diffusion. To solve these two problems, this article introduces a new method which contains a new network architecture. The new architecture has a parallel progressively complementary structure, which is used to separate content reconstruction and structure correction, thus reducing the burden of the decoder. In addition, the new architecture is embedded with correction layers in skip connection, aiming to pre-correct the images before transferring with the help of the appearance flows in different layers. As a result, the decoder can readily rebuild a credible result with the less distorted features. This new method shows high superiority according to different experimental results.

Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: CHEN Siyu / CPEG Course: UROP1100, Fall UROP2100, Spring

This paper presents a summary of the two main algorithms we used for detecting surrounding buildings or sites and their practical implementation process through python. Mobile applications need to obtain information about surrounding buildings and sites before pervasive positioning. How to determine the required buildings and sites are implemented in cloud server side. In this paper, we surveyed the related work in this field and concluded two algorithms for detecting surrounding buildings or sites with some input parameters. We also implemented them into practice python code to validate their correctness. We believe this paper can make a relatively comprehensive summary and make contribution to further implementation of the project.

97

School of Engineering Department of Computer Science and Engineering

Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: DING Yiyi / COMP Course: UROP1100, Spring UROP2100, Summer

In the past few years, indoor localization technologies have experienced rapid development. Specifically, proximity-based and fingerprint-based localization can precisely find out user location at a fast speed. In the outdoor localization aspect, the Global Navigation Satellite System (GNSS) has been applied for many years and is able to provide precise localization results to the users. In this case, current technologies are mature enough to support pervasive positioning, which is to locate an object anywhere seamlessly on a country scale. However, different techniques and data formats are used in these localization systems, and they are not transferable. These mean that a standard that can interconnect these positioning techniques together should be developed in order to achieve pervasive positioning,

Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: GUO Zhi / ISDN Course: UROP1100, Fall

Continuing the project in summer UROP1000, I have been doing the UROP1100 project: Indoor Localization and Mobile Computing by Professor CHAN Gary Shueng Han in the fall semester 21-22. Further study from previous work, I tried to read more difficult papers during the three months and gained much more valuable knowledge than in UROP1000. Those papers offered me a brand-new world which is totally different from my previous understanding. I obtained a much deeper understanding of the indoor localization technic.

98

School of Engineering Department of Computer Science and Engineering

Deep Learning for Ophthalmology Image Analysis Supervisor: CHEN Hao / CSE Student: WANG Yueying / DSCT Course: UROP1100, Fall UROP2100, Spring

Traditionally, Diabetic macular ischaemia (DMI) has been defined and graded on the angiographic basis of the enlarged and irregular avascular area of the fovea. Using advanced optical coherence tomography angiography (OCTA) to identify early DMI phenotypes is the key to successful treatment of DMI. However, due to the patient-related artifacts, technically related artifacts or both related artifacts, the judgment result of deep learning models is not reasonable. In this undergraduate research opportunity project, we tried to add prior knowledge to the model according to auxiliary learning and multi-task learning, and relabeling the dataset to achieve better results.

Deep Learning for Ophthalmology Image Analysis Supervisor: CHEN Hao / CSE Student: ZHANG Weiwen / DSCT Course: UROP2100, Fall

Optical coherence tomographic angiography (OCTA) is a novel non-invasive technology that can capture retinal and choroidal microvascular images without any invasive operation. The scan region of OCTA is typically 3×3-mm and 6×6-mm. Although 6×6-mm has a larger field of view, its defect is the relatively low resolution compared with 3×3-mm due to under-sampling. This paper first proposed an unsupervised highresolution reconstruction pipeline in the frequency domain based on the structure of CycleGAN and made progress comparing it with vanilla CycleGAN. Based on it, this report continued to propose a network that does not need interpolation algorithms as preprocessing and constrains the training in the frequency domain. This paper has been published in top-tier conference MICCAI: https://link.springer.com/chapter/10.1007/978-3-031-16434-7_62

99

School of Engineering Department of Computer Science and Engineering

Machine Learning in Healthcare Supervisor: CHEN Hao / CSE Student: KHAN Asif / COMP Course: UROP1100, Spring

In light of the recent developments in artificial intelligence and medical technologies, researchers all around the world have been exploring and implementing various machine learning methods geared toward healthcare applications. A field that is most commonly researched is medical image processing. Machine learning can be utilized by researchers and doctors to optimize and provide more accurate diagnoses through medical image analysis, and several architectures and algorithms have been designed to facilitate effective image processing techniques.

An initial starting point for further boosting medical research would be to start through the method of nuclei detection. With the help of the datasets provided by Kaggle’s 2018 Data Science Bowl, this report would build and evaluate several networks for image segmentation. This report would also explore the potential prospects of machine learning and how it could effectively be used in healthcare applications.

Machine Learning in Healthcare Supervisor: CHEN Hao / CSE Student: SINGH Har Shwinder / PHYS Course: UROP1100, Fall

This paper proposes a Deeply Supervised Attention U-Net Deep Learning network with a novel image mining augmentation method to segment brain tumors in MR images. The network was trained on the 3D segmentation task of the BraTS2021 Challenge Task 1. The Attention U-Net model improves upon the original U-Net by increasing focus on relevant feature maps, increasing training efficiency and increasing model performance. Notably, a novel data augmentation technique termed Positive Mining was applied. This technique crops out randomly scaled, positively labelled training samples and adds them to the training pipeline. This can effectively increase the discriminative ability of the Network to identify a tumor and use tumor featurespecific attention maps. The metrics used to train and validate the network were the Dice coefficient and the Hausdorff metric. The best performance on the online final dataset with the aforementioned network and augmentation technique was: Dice Scores of 0.858, 0.869 and 0.913 and Hausdorff Distance of 12.7, 16.9 and 5.43 for the Enhancing Tumor (ET), Tumor Core (TC) and Whole Tumor (WT).

100

School of Engineering Department of Computer Science and Engineering

Deep Learning for Medical Image Analysis Supervisor: CHEN Hao / CSE Student: DENG Yufan / SENG Course: UROP1000, Summer

Deep-learning-based algorithms have done well in helping detect diseases from various medical images. However, medical image data to support large-scale model training are scarce and data augmentation methods for normal images fail to preserve medical meaning, which is extremely important. We proposed a synthetic data augmentation method that can preserve meaningful attributes of medical images using disentangled representation. Our method first uses encoders to extract domain-invariant features and domain-specific features between medical images of different qualities. Then, Generative Adversarial Networks (GANs) are used to generate new images that only keep the domain-invariant feature but with attributes from a new domain. Visualization of augmented data shows our method can preserve important feature of original images, and experiments on disease grading show our method have the potential to help improve grading accuracy. In the future, we will keep improving the method and evaluating it with more medical tasks.

Deep Learning for Medical Image Analysis Supervisor: CHEN Hao / CSE Student: GUO Meichen / IPO Course: UROP1000, Summer

Machine Learning has become a hot topic in high-tech world, with applications in various fields, especially in medical care. Core to many of these applications are visual recognition tasks such as image classification, localization and detection. Through the Undergraduate Research Opportunities Program, under the guidance of Professor Hao CHEN and Doctor Luyang LUO, I went deep into the field of deep learning. Specifically, I first studied some online materials that give a brief introduction to deep learning and machine learning. Then, I was assigned to read two papers on DualConsistency Semi-Supervised Learning and TransUNet, and try to run the relevant codes. I am writing this paper as a reflection.

101

School of Engineering Department of Computer Science and Engineering

Deep Learning for Medical Image Analysis Supervisor: CHEN Hao / CSE Student: LAU Ying Yee Ava / DSCT Course: UROP2100, Fall

This project aims to analyse the health conditions of the patella of patients, so as to determine the type of surgery, either total knee arthroplasty (TKA) or unicompartmental knee arthroplasty (UKA), is required. Given slices of 2D binary segmentation masks of the patella, it is reconstructed into a 3D triangular mesh, then several descriptors are computed based on the features of the mesh. These descriptors are later inputted to the logistic regression model, which in the end provides a probability for having TKA or UKA. The predictive ability of the logistic regression model is examined: it could provide predictions with accuracy around 80% when compared to the ground truth.

Deep Learning for Medical Image Analysis Supervisor: CHEN Hao / CSE Student: WIOGO Calvin / SENG Course: UROP1000, Summer

The Coronavirus disease (Covid-19) has resulted in millions of deaths and is still posing a considerable risk to the public’s well-being, both physically and mentally. The development of automatic segmentation has mitigated this situation by enabling medical professionals to perform screening, treatment planning, and follow-up assessments effectively on infected patients. This research paper proposes TransUNet, a method of processing Two-Dimensional CT images for Covid-19 segmentation. TransUNet comprises U-Net (a form of Convoluted Neural Network) and transformer. U-Net has become prevalent in the medical field for medical image segmentation and its implementation has shown great success on multiple medical occasions, namely Magnetic Resonance (MR) Cardiac Segmentation and Computed Tomography (CT) Organ Segmentation. The transformer employs a self-attention mechanism to model long-range relations, overcoming the limitations of the Convolution Neural Network having local sensitivity.

102

School of Engineering Department of Computer Science and Engineering

Deep Learning for Medical Image Analysis Supervisor: CHEN Hao / CSE Student: ZHOU Taichang / COMP Course: UROP1100, Summer

The explainable AI is artificial intelligence (AI) in which humans can understand the decisions or predictions made by the AI, which uncovers the 'black box' in the deep learning model. As deep learning models are widely used these days, those models do make sense in many areas. However, for some specific area, i.e., medical image analysis and financial investment, the requirement for explainable models are growing because of their severe outcome caused by mistakes, if any.

Here, we will introduce some common post-hoc methods for CNN models used in XAI, including both gradient based and non-gradient based, local and global.

Data-Efficient, Domain Generalizable and Interpretable Deep Learning Supervisor: CHEN Hao / CSE Student: CHENG Yuhan / COSC Course: UROP1000, Summer

Deep learning generates a computing model by constructing an intricate neural network. It requires a mass of data to train. However, it is impossible to visit all domains of data so the model cannot work well enough. Domain generalization (DG) was born to solve this issue. It is aimed at achieving generalization for a deep learning model in the unseen dataset by using available source data. DG has been applied widely in computer vision, medical image analysis, natural language processing, and so on. In this project, we mainly focus on medical image analysis, especially diabetic retinopathy images. This report presents some basic knowledge of domain generalization and introduces several commonly used deep learning methods. In addition, this report records my learning progress in this project during this summer and some following work that needs to be done.

103

School of Engineering Department of Computer Science and Engineering

Data-Efficient, Domain Generalizable and Interpretable Deep Learning Supervisor: CHEN Hao / CSE Student: WANG Zeyu / COMP Course: UROP1100, Fall

Deep learning models have shown great potential in detecting mitotic figures in histopathology images, the density of which is known to be highly related to tumor proliferation and thus crucial in diagnosing cancer. However, performance of a model would degrade significantly if images for testing are from different laboratories than the images for training. This is caused by domain gaps due to difference in specimen acquisition methods. Under hypothesis that the whole slide image(WSI) scanner device plays a decisive role, the MICCAI-MIDOG challenge is organized to seek for machine learning solutions that are invariant to the domain shift. In this UROP project, I developed augmentation-adversarial-based pipeline to resolve the domain gaps.

Data-Efficient, Domain Generalizable and Interpretable Deep Learning Supervisor: CHEN Hao / CSE Student: ZHANG Yuhao / SENG Course: UROP1000, Summer

Diabetic Retinopathy (DR), often occurs with diabetes mellitus. It is a leading cause of blindness for many patients. Automatic grading may help the patient receive the healing earlier. This report will study and discuss several different ways of using machine learning to grade Diabetic Retinopathy automatically. Among all those models, I discovered that the Convolutional Attention Block plays an important role in grading Diabetic Retinopathy. By using several different models which used the Convolutional Attention Block and changing the hyperparameters, this report tries to discover a way to grade Diabetic Retinopathy by using machine learning methods that reaches a high accuracy.

104

School of Engineering Department of Computer Science and Engineering

Data-Efficient, Domain Generalizable and Interpretable Deep Learning Supervisor: CHEN Hao / CSE Student: ZHU Zhengjie / COMP Course: UROP1000, Summer

Due to certain constraints between ethics and the specificity of pathology slices, the datasets on digital pathology are not very large. Moreover, there are often only some tiny or local datasets instead of big public datasets like ImageNet. These images were taken in different ways, with different coloring techniques and resolutions, making applying deep learning to pathology more challenging. Our goal is to train a generalized pre-trained model for digital pathology. Then, others can apply it to downstream tasks using a small amount of data to fine-tune it to get good results. Also, self-supervised training has proven its effectiveness in recent years. It is also well adapted to the problem that most datasets are not labeled, so I researched many self-supervised training models.

Deep Video Super-resolution Supervisor: CHEN Qifeng / CSE Student: CHENG Yuhan / COSC Course: UROP1100, Fall

Video compression is the process of encoding video information into numerical signal using fewer bits. Since videos could only be transmitted and stored by bits, so video compression is necessary to reduce the size of a video file. Video compression algorithm is widely used in all fields involving videos. A better algorithm means better quality, faster transmission rate and smaller size. All compression algorithms can be classified into two kinds, lossy and lossless. The video quality would be affected in lossy video compression algorithm. But no information is lost in lossless compression. An integrated video compression standard involves both lossy and lossless compression. There are many organizations which sets video compression standards. But two of them are the most influential. One is International Telecommunications Union – Telecommunication Standardization Sector (ITU-T). Its subordinate department VECG has set many important video compression standards such as H.261, H.263 and H.264. The other one is International Standards Organization. MPEG-1 and MPEG-4 are set by its subordinate department MPEG.

105

School of Engineering Department of Computer Science and Engineering

Deep Video Super-resolution Supervisor: CHEN Qifeng / CSE Student: HU Wenbin / COMP Course: UROP1100, Summer

Autonomous driving (AD) can significantly change people’s life. It will emancipate the labor force and save many resources by reducing the number of vehicles and travel time because cars can be scheduled by computers. Due to the promising future of autonomous driving, I have a keen interest in AD works, so I surveyed some works about AD in this summer UROP project and will keep focusing on AD in the future. AD’s work pipeline encompasses several parts: Maps and Sensors, Perception, Prediction, Planning, and Control, as shown in figure 1. In this UROP project, I mainly focused on motion forecasting for self-driving vehicles.

Map & sensor perception prediction planning control

Figure 1. Autonomous driving working pipeline.

Deep Video Super-resolution Supervisor: CHEN Qifeng / CSE Student: LIN Siyan / SENG WANG Zixuan / SENG ZHANG Liyu / SENG Course: UROP1100, Summer UROP1100, Summer UROP1100, Summer

Image denoising has been a heated topic in image restoration for a long time, and the convolutional neural network (CNN) has attracted considerable attention due to its favorable denoising performance. In this report, we trained and evaluated three different CNN based models on their denoising performance, namely the ResNet, the UNet, and the Noise2Void. We used the same dataset for evaluation and compared their statistical results for denoising. Our results showed that ResNet performed best especially when the noise level is low, while UNet and Noise2Void are more stable. From this perspective, we reflected on the issues we encountered and proposed possible directions for further improvement.

106

School of Engineering Department of Computer Science and Engineering

Deep Video Super-resolution Supervisor: CHEN Qifeng / CSE Student: VUONG Tung Duong / DSCT Course: UROP1100, Spring

The Super-Resolution task in video using deep networks is the main topic of my UROP project. This term, we focus on videos of human faces, which have various application. Using low resolution video, we hope to improve its quality efficiency by a number of idea. The first approach is to use NeRF to learn the 3D presentation of the object and increase the resolution by align the textures from some highresolution images. The second approach is to use C2-Matching, a method that try to match the correspondences between lower images and high resolution reference. Both approach can be applied together for a better result. Moreover, there is room for improvement in both approach.

Deep Video Super-resolution Supervisor: CHEN Qifeng / CSE Student: ZHANG Hanning / COSC Course: UROP1100, Summer

I participated in a project called Stereo Audio Localization in July and August, during which I collaborated with my schoolmate Huihao, supervised by professor and graduate students. The project’s goal is predicting the location on a reference image based on Stereo footstep. Our mission is to expand the dataset, making it more diverse, and try to improve the machine learning if possible. At the same time, I also spend much time learning how to read through and run a project code of the paper, how to modify the project to meet different requirements. Until now, we have collected totally 37 single person video and 21 double-person video, with on campus and off campus environments. And we have tried to train part of our samples.

107

School of Engineering Department of Computer Science and Engineering

Deep Video Super-resolution Supervisor: CHEN Qifeng / CSE Student: ZHANG Juntao / COSC Course: UROP2100, Summer

Scene text image super-resolution aims to enhance the readability of low-resolution scene text images by increasing their original resolution. It can be used as a pre-processing step in scene text image recognition or to assist human readers. By learning end-to-end using paired high-resolution (HR)-low-resolution (LR) images, a scene text super-resolution model can significantly improve the accuracy of text recognition models of low-quality images and super-resolve human-unreadable blurred document images. In this paper, we proposed a Prior-Guided Super-Resolution (PGSR) structure to address this problem, which was inspired by the Text Attention network (TATT) . Furthermore, we proposed an instance of PGSR, hybridizing two state-of-the-art models, TATT and Scene Text Telescope (STT). To test the performance, we evaluate the output by traditional image super-resolution PSNR/SSIM metrics, as well as the recognition accuracy of text recognition models like CRNN. Code is available at GitHub.

Neural Rendering Supervisor: CHEN Qifeng / CSE Student: CHENG Yize / COSC Course: UROP1100, Summer

Autonomous Driving (AD) is an emerging field that is growing rapidly. This has been particularly the case since the development of deep learning architectures, such as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN) and Graph Neural Networks (GNN). Common tasks for autonomous driving applications include object detection and tracking in videos, motion trajectory prediction and trajectory planning. In this UROP project, we studied the paper Learning Lane Graph Representations for Motion Forecasting. The proposed LaneGCN architecture for motion forecasting is still very commonly used in trajectory prediction tasks in current autonomous driving applications. In this report, we try to unscramble this paper in our words based on our own understanding and make the ideas more comprehensible for beginners than the presentation in the original paper.

108

School of Engineering Department of Computer Science and Engineering

Neural Rendering Supervisor: CHEN Qifeng / CSE Student: KIM Hyeonjae / COSC Course: UROP1100, Spring

In a recent couple of years, a variety of methodologies to represent the surrounding scenes in a continuous manner have been proposed. Especially, neural-based implicit representation has received tremendous attention due to its potential impact on super-resolution and other downstream tasks. In this work, we specifically focus on learning a non-discrete (i.e., continuous) representation of images using multilayer perceptron (MLPs) with nontrivial activation layers. To be specific, influenced by the recent research on the implicit neural representation of images, we propose a way to improve the Local Implicit Image Function (LIIF) by adopting the Sinusoidal Representation Networks (SIREN). By experimenting with the CelebAHQ dataset, we show this simple modification leads to significant improvement in performance both in qualitative and quantitative ways.

Neural Rendering Supervisor: CHEN Qifeng / CSE Student: PHAM Trung Kien / DSCT TOH Magdalene Youjun / COMP ZHAO Jiachen / COSC Course: UROP1100, Spring UROP1100, Spring UROP1100, Spring

Video inpainting by internal learning is predicting the texture of masked region of a sequence of frames of a video only depending on the coherence and consistency among frames. Our goal is to improve the performance of inpainting model working on a special type of video, stereo video. To achieve this goal, new losses, named disparity loss and edge loss, are proposed and effective methods, such as mask augmentation, are used. Our model is trained on different segments of stereo video and obtains visually pleasing results. Compared to previous model trained only on single view side, our model trained on both view sides outperforms and the improvement is satisfying.

109

School of Engineering Department of Computer Science and Engineering

Geodesic Paths in Meshes Using Edge Flips Supervisor: CHENG Siu Wing / CSE Student: GAO Zhimeng / COMP Course: UROP1100, Fall

This project concerns the study of a new method of finding Geodesic Paths in Meshes Using Edge Flips. We learned some pre-knowledge of this new method about geometry, which involves reading some reference books and doing practice problems. After that, we went through some tutorials about Github and tried to run the author’s code of this method in our local environment. Also, we plan to carry out an implementation to verify the practical effectiveness and experiment to see how the method works for graphs in general.

Automated Program Synthesis Supervisor: CHEUNG Shing Chi / CSE Student: WANG Yiran / QFIN Course: UROP1100, Spring UROP2100, Summer

Programs can be synthesized automatically by methods such as programming-by-examples. Through example pairs of code inputs and outputs inputted by users, we could achieve automation in program generation. And in this project, we focus on program analysis so developers would not need to write the scripts repeatedly for code analysis purposes on different projects. We collected benchmarks from GitHub, research papers at the International Conference on Software Engineering (ICSE) and Foundations of Software Engineering (FSE). We gathered all their ground-truth analysis programs and designed a domain-specific language as mentioned in the previous report. Finally, we provided users a website interface for inputting their input-output examples.

110

School of Engineering Department of Computer Science and Engineering

Detecting Deep Learning Software Defects Supervisor: CHEUNG Shing Chi / CSE Student: LIU Jianmeng / COMP Course: UROP1100, Spring UROP2100, Summer

Deep learning (DL) systems have already been an essential component in our life, based on which deep learning applications are built. Nevertheless, limited work has been proposed on testing the underlying deep learning libraries, such as the popular PyTorch and TensorFlow, which are the foundations of DL systems. In our previous work, we collected test oracle changes of unit tests in TensorFlow, and explored the reasons and characteristics of the modifications through empirical studies. During this stage, we conducted another study to test the strength of unit tests provided by TensorFlow developers, where we tried to modify the data type or randomize the values of test inputs. Results showed that many tests would fail due to different reasons. Based on this study, we consider that performing fuzzing in deep learning libraries may be able to detect defects effectively. We examined FreeFuzz, a previous work on fuzzing DL libraries, and our preliminary results show that by increasing the volume of input sources from open source, the varieties of both test inputs and methods under test can be greatly increased, suggesting that this might bring huge improvements on the effectiveness of fuzzing.

Detecting Deep Learning Software Defects Supervisor: CHEUNG Shing Chi / CSE Student: TANG Zexi / MATH-SFM Course: UROP1100, Spring

Software defect detection is generally used for distinguishing the deficient modules which exist in the software so that it can help in improving the quality of the software application. Machine learning techniques, more specifically, deep learning tactics, are useful in terms of software defect detection. Software has two categories, defective and non-defective modules. Decision Tree and Logistic Regression methodologies are developed for the classification of defective data set. The datasets are taken from the reliable data repository thus consistency and accuracy is guaranteed. During this UROP process, we study about all kinds of algorithms and code implementation about detecting deep learning software defects and collecting data from the online discussion platform related to common pytorch issues.

111

School of Engineering Department of Computer Science and Engineering

Building a Blockchain and Smart Contract Application Supervisor: CHEUNG Shing Chi / CSE Student: NGUYEN Kha Nhat Long / COMP Course: UROP1100, Spring

Blockchain has become very popular and outstanding technology in recent day. The technology helps us to deal with the trust problem that occur for many years. With a large amount of user, security becomes a big problem for blockchain since it directly relates to money. With the evolution of blockchain, people are able to do many thing on its and the we can write more complicated code by using smart contract. This lead to a problem that many bug has happened on the blockchain and cause lost of money. Example can see that the DAO attack lost 80 million dollar. The important is to find the bug on smart contract by using an general method. I will give some investigation on two popular attack on block chain: On chain off chain and front running.

Building a Blockchain and Smart Contract Application Supervisor: CHEUNG Shing Chi / CSE Student: PHAM Thanh Lam / MATH-CS Course: UROP1100, Spring

Smart contract auditing is critical because once a smart contract is published to the main network of the blockchain system, it cannot be modified and remains in effect indefinitely. All the contract's flaws and vulnerabilities will be exploited, resulting in massive financial and token losses. In our present work, developing a tool that can be utilized for several purposes, such as visualizing a transaction, analyzing a smart contract, and so on, necessitates the use of smart contract analysis tools. As a result, developing a parser to transform the output of such tools into the appropriate format is a critical effort. My current emphasis is on the ETHRACER tool, which detects Event-Ordering bugs (EO bugs). This report helps understanding EO bugs, front-running attack, how ETHRACER works in brief, and how I came up with a parser.

112

School of Engineering Department of Computer Science and Engineering

Building a Blockchain and Smart Contract Application Supervisor: CHEUNG Shing Chi / CSE Student: SON Hangyul / CPEG Course: UROP1100, Spring

As the blockchain ecosystem grows, the amount of hacked cryptocurrencies has significantly grown over the years. Based on 2022 Crypto Crime Report, total sum of 14 billion us dollars have been received by illicit addresses in 2021, which nearly doubled from 7.88 billion USD in 2020. In order to fully adopt and enjoy the benefits that blockchain technology can bring to the world, the security guarantees must be held. Therefore, a phd candidate, Wuqi ZHANG, is conducting a research to propose a new approach to analyze smart contracts of Ethereum Blockchain. The existing security analysis tools each face various limitations, unable to fully safeguard against the historical attacks. A benchmark must exist so to quantify the performance of the existing tools in comparison to the new approach, which is yet to be implemented. This project focuses on implementing tools which would be used to create a benchmark based on historical attacks that occurred on the Ethereum Blockchain.

Building a Blockchain and Smart Contract Application Supervisor: CHEUNG Shing Chi / CSE Student: YU Chi Kai / QSA Course: UROP1000, Summer

Event-ordering bugs (EO bugs) are intimately associated with the dynamic ordering of contract events. The past techniques to detect EO bugs have been restricted to bugs which just involve 1 or 2 event orderings. Ethracer could runs directly on Ethereum bytecode without the necessity of source code to detect EO bugs automatically. Due to limited time because of internship and GRE preparation, I apologized that I did not have enough time to explore Ethracer’s branch coverage on dynamic symbolic execution. However, I did try to run Ethracer on the Ethereum smart contracts to see its performance and limitations. I would briefly summarized what I have learnt in this UROP. Files could be found in my Github.

113

School of Engineering Department of Computer Science and Engineering

Analysis of Hierarchical Clustering Methods Supervisor: GOLIN Mordecai Jay / CSE Student: RUBAB Tamzid Morshed / DSCT Course: UROP1100, Fall

In this report, I analyze two hierarchical clustering algorithms for graphs drawn from planted partition model. In this model, nodes are divided into k clusters at first, and intra-cluster edges are added independently with some probability p and inter-cluster edges are added independently with probability q, where p > q. For the first algorithm, we assume that the partitions (clusters) are known. In this case, we can assume a generalized version of planted partition (intra-cluster probabilities can be anything larger than q). And for the second one, the partitions are unknown. The former achieves 3k−2 3k approximation and the latter asymptotically finds at least 0.5-approximation.

Constructing AIFV Codes Supervisor: GOLIN Mordecai Jay / CSE Student: ZHAO Jiachen / COSC Course: UROP1100, Summer

Almost Instantaneous Fixed to Variable 𝑚𝑚 bit delay coding, also known as AIFV-m coding, is a lossless compression method for memoryless sources. Compared to Huffman coding, it sacrifices instantaneity for better performance. Instead of constructing a single tree like Huffman coding, AIFV-m coding generates 𝑚𝑚 trees to encode and decode the text. To construct the optimal 𝑚𝑚 trees, both local optimization, aiming to build the optimal m-th tree given certain conditions, and global optimization, aiming to coordinate different trees to achieve optimal performance, are necessary. This paper introduces an implementation of a new local optimization procedure for AIFV-m coding proposed in [2], which runs in 𝑂𝑂(𝑚𝑛𝑚+2) time and 𝑂𝑂(𝑚𝑛𝑚+2) space.

[2] Mordecai J. Golin and Albert John L. Patupat. “Speeding Up AIFV-m Dynamic Programs by m–1 Orders of Magnitude.” 2022 IEEE International Symposium on Information Theory (ISIT’22). 2022.

114

School of Engineering Department of Computer Science and Engineering

Machine Learning Driven Web Log Tracing System Supervisor: IEONG Sze Chung Ricci / CSE Student: LEI Yongyu / CPEG Course: UROP1100, Fall

Network Intrusion Detection System (NIDS), as the system aiming to detect intrusion by monitoring network activities, has become one critical part of cybersecurity. Web logs generated from network traffic of a system are mostly used as the objective of monitoring in a NIDS. Along with the prevalence of artificial intelligence (AI), to improve the performance of NIDSs, machine learning (ML) and deep learning (DL) methods have been widely applied to NIDSs to improve the accuracy of intrusion detection in recent three decades. This report trained and tested different ML or DL model based NIDSs using UNSW-NB15 dataset and summarized their performances of classifying network traffic records which would provide essential references for further proceeding the project.

Optimization Problems in Blockchain Ecosystems Supervisor: KAFSHDAR GOHARSHADY Amir / CSE Student: ZHU Zherui / COGBM Course: UROP1100, Fall

Fall Semester of 2021 is my first semester with professor Amir to do the UROP project: Optimization Problems in Blockchain Ecosystems, my learning target in this semester is mainly to learn knowledge related to the project focus: 'finding more efficient algorithms for various optimization problems that are currently being solved inefficiently or imprecisely in the day-to-day operation of blockchains.' The areas I have covered include the basic understanding of Blockchain. Understand algorithm optimization, such as parameterized algorithm. Then I read the Paper: Optimal Mining: Maximizing Bitcoin Miners’ Revenues which is closely related to the project focus. Finally, I chose A topic: A Survey of Attacks on Ethereum Smart Contracts (SoK) for study. This report will show my learning progress and achievement.

115

School of Engineering Department of Computer Science and Engineering

Parameterized Algorithms for Static Program Analysis Supervisor: KAFSHDAR GOHARSHADY Amir / CSE Student: CHENG Tien-chun / COMP Course: UROP2100, Fall UROP3100, Spring

Data locality problem is an important topic in program optimization, and the hierarchical model is one of a way we can cluster data. Previous studies have proved the hardness of a task associated with constructing the hierarchical model. In this report, we attempt to find a polynomial-time algorithm parameterized by treewidth. In light of the difficulty in finding a direct parameterized algorithm, we discuss a side approach to this problem and propose some properties about treewidth. At the end of the report, we discuss how we might utilize these properties in finding a parameterized algorithm and propose possible future work.

Parameterized Algorithms for Static Program Analysis Supervisor: KAFSHDAR GOHARSHADY Amir / CSE Student: HUANG Yi-feng / COMP Course: UROP1100, Fall

Static program analysis aims to evaluate the program through abstraction of the program and provide useful information about the program. Useful examples include live variables analysis, available variable expressions analysis, and flow analysis. These analyses help with improving the efficiency of the program or help verify the correctness of the program.

Many of the analysis in static program analysis can be modeled by a graph. And we had learnt parameterized algorithm for graph with bounded treewidth and path width such that certain order of traversing the graph can be exploited to reduce the time complexity of the algorithm running on the graph. In this research project, we studied the parameterized algorithms and try to use parameterized algorithm in solving problems in static program analysis, and hopefully on the area of points-to analysis, which analyze the point to source and target relationship.

116

School of Engineering Department of Computer Science and Engineering

Parameterized Algorithms for Static Program Analysis Supervisor: KAFSHDAR GOHARSHADY Amir / CSE Student: TSAI Yun Chen / MATH-IRE Course: UROP1100, Spring UROP2100, Summer

In this semester we focus on problems related to language inclusion and shortest lasso example problem, these problems are closely related to the application in model checking. In particular, most systems in real life are of low treewidth, thus parameterized algorithm over treewidth will be useful in performing practical model model checking. This report summarize two main pieces of literature that we focus on during the semester, major contents are captured and presented. Further we discussed about the possible application and future research direction at the end of the report.

Assess User Experience to Design Effective Visual Representation and Interaction in Virtual Reality Supervisor: MA Xiaojuan / CSE Student: DOU Peihao / COSC Course: UROP1100, Summer

This summer, I participated Urop1100 project with title ‘Assess user experience to design effective visual representation and interaction in virtual reality’. I have experienced VR before but never really went into it as a developer. It takes some time and efforts to learn the use of the VR headset and to create a VR app by myself. This report mainly focuses on my learning progress, which includes writing small applications using unity and importing it to VR Headset, and user experience to some developing VR apps and also hand interface designs.

Assess User Experience to Design Effective Visual Representation and Interaction in Virtual Reality Supervisor: MA Xiaojuan / CSE Student: JIA Feiyu / DSCT Course: UROP1100, Summer

Virtual Reality, short for VR, is a developing research direction in the field of digital technology. It has drawn both interest from citizens and entrepreneurs, as a novelty that may change their life. As the number of VR users is growing up, the design of effective visual representation and interaction is more and more important for VR developers. VR is outstanding to join people in the fancy scenery and inaccessible experience without time and space limitations. However, its dizziness and operation difficulty keep a lot of people away from its unique fascination. This project will show the process to learn to program VR in Unity and improve the user experience by designing effective visual representation and interaction.

117

School of Engineering Department of Computer Science and Engineering

Assess User Experience to Design Effective Visual Representation and Interaction in Virtual Reality Supervisor: MA Xiaojuan / CSE Student: XU Zian / DSCT Course: UROP1100, Summer

With the development of immersive technology, virtual reality (VR) devices have become more and more accessible to normal people. We want to explore more potential uses of virtual world applications, especially their educational use through telling a story. Besides, people have paid more and more attention to public health in recent years, especially under the covid-19. In this project, we design a VR story about public health in a classroom. And based on this model, we want to understand how users feel about different data representations in VR and how users’ interaction with VR affects their understanding of the story.

Predicting Student Performance on an E-learning Platform Supervisor: MA Xiaojuan / CSE Student: KIM Justin Juho / MAEC Course: UROP1100, Summer

This paper aims to fill knowledge gaps on the role of peer learning in massive open online courses (MOOCs); chiefly, the ways in which peers interact with one another in MOOCs (RQ1), the factors inhibiting peer learning (RQ2), and interventions in the instructional design of MOOCs to better facilitate peer interaction and learning (RQ3). Semi-structured interviews were conducted and coded. It was found that MOOC learners used forums, social media, and messaging platforms to interact with one another, of which forums were the least conducive to peer interaction. Furthermore, respondents had reservations about the value of peer evaluation, eroding its effectiveness as an evaluation technique. Overall, this study presents three recommendations regarding the instructional design of MOOCs: (i) establish and encourage informal methods of communication between MOOC learners, (ii) assign set partners to learners for the duration of the MOOC, and (iii) limit peer evaluations and peer-graded assignments to the final weeks of a MOOC.

118

School of Engineering Department of Computer Science and Engineering

Interaction Design For Human-AI Collaboration Supervisor: MA Xiaojuan / CSE Student: YAO Chongchong / DSCT Course: UROP1100, Summer

The game of Go, despite thousands of years of history and numerous legendary players, is currently under the absolute dominance of artificial intelligence (AI). Motivated by such situations, a considerable amount of professional Go players have started to improve their skills by playing against AI and studying AI’s moves. However, few Go software allows humans and AI to collaborate as mates. In this study, we developed a Go software that supports human-AI collaboration (HAIC), aiming to facilitate AI-assisted training. To enhance the user experience, the GUI of the software was modeled after KaTrain, a popular open-source Go tool. In addition, the software features a calibrated AI assistant able to adjust its skill level according to the user’s capability.

Develop a Machine Learning Application on Video Data Base Retrieval Using Tensorflow Facilities Supervisor: NG Wilfred Siu Hung / CSE Student: CHO Heejin / DSCT Course: UROP1100, Fall

This essay explores and develops a machine learning application on video data base retrieval using Tensorflow facilities. To better develop the project, the basic definitions and descriptions of machine learning, tensorflow, keras will be studied. Further, data collection and the methodology of how the video classification model can be built will be presented. While developing ideas and meaningful feedback from the suprevisor, it was worthwhile to clarify the steps while learning concepts accurately and precisely to enrich fundamental knowledge in depth. Hence, the aim of the topic was not partially achieved. However, it was a great opportunity for me to enhance soft skills on machine learning and I hope to continue this topic in future.

119

School of Engineering Department of Computer Science and Engineering

Developing Tools for Privacy-preserving Financial Applications Supervisor: PAPADOPOULOS Dimitris / CSE Student: GU Yi / COMP Course: UROP1100, Spring UROP2100, Summer

The UROP Program “Developing tools for privacy-preserving financial applications” introduces a blockchainbased data market application. Based on the spring semester’s study, the student finished the design of the Zero Knowledge protocol in the Agora application, the implementation of the decryption function, and the interaction with the smart contract. The application can now connect to the Ethereum network and perform Agora’s basic functions. The data consumer can verify the correctness of the data from the data broker without revealing any useful information to other third parties in the same blockchain. Moreover, all the parties involved in the data transactions will rely on Ethereum’s smart contracts to complete part of the functionalities and data storage.

Developing Tools for Privacy-preserving Financial Applications Supervisor: PAPADOPOULOS Dimitris / CSE Student: WU Lijia / COMP Course: UROP1100, Fall

In financial auditing, companies need to provide all bookkeeping records for the government or auditing firms to trace the transactions. However, it incurs higher trust on the auditors. Collections of the bookkeeping records can also be tampered. Permissioned blockchain is a distributed ledger technology that can provide the service of confidentiality and scalability. With cross-chain communication, transaction can be automatically recorded for auditing. In this project, we have implemented two smart contracts, bookkeeping and cross-chain on Hyperledger Fabric to build a 2-tier permissioned blockchain that provides fundamental services for financial auditing, and evaluated its performance in different platforms.

120

School of Engineering Department of Computer Science and Engineering

Compilation, Optimization, and Metaprogramming Supervisor: PARREAUX Lionel / CSE Student: QIU Jiangkun / ISDN Course: UROP1100, Spring

As developers, we seek the triangular balance of writing fast, reading fast, and running fast. Abstraction is an excellent way to achieve this, as we abstract and encapsulate well-written algorithms and optimizations into functions, classes, and libraries. However, many abstractions come with a penalty. Many languages state that they have zero-cost abstraction, but are they really zero-cost? Compilers fail to perform some optimizations behind abstractions and pay a runtime performance cost. Even as simple as invoking a function without inlining has some runtime cost and may influence CPU cache and branch predictions; generics in GoLang have runtime performance penalty; Java relies on virtual tables to invoke methods once failed to infer the specific method. This leaves space for our Programmable Zero-Cost Abstractions, which heavily utilized macro, comp-time interpretation, and specialization. We can avoid many abstraction costs at runtime by leaving more work to comp-time.

Compilation, Optimization, and Metaprogramming Supervisor: PARREAUX Lionel / CSE Student: YUAN Keyu / COMP Course: UROP1000, Summer

Specialization is a useful technique for program optimization, and on top of that, in PZCA(Programmable Zero-cost Abstractions), program monomorphization is implemented to produce better optimization results. In this UROP we implemented a regular expression matcher as the front end of PZCA, which demonstrated the power of this program optimizer. The matcher takes in a static regex as a class hierarchy and a dynamic string, and applies a naive matching function to determine whether the string matches the regex. The original matching function was slow, but after adequately specialized, the algorithm become as fast as compiling regex.

121

School of Engineering Department of Computer Science and Engineering

Programming Language Design and Implementation Supervisor: PARREAUX Lionel / CSE Student: LUO Yicong / COMP Course: UROP1100, Spring

During this UROP course, I mainly contribute to the development of the mlscript, an object oriented functional language with principal type inference developed by Prof Lionel Parreaux. The first feature added to the project is the array type, which will be the super type to the existing tuple. The second one is the addition of mutable fields for record types. The last contribution to the project during the UROP period is the splice type, which is still in progress. This report will describe and summarize all the work and contribution to the mlscript language and future works.

Programming Language Design and Implementation Supervisor: PARREAUX Lionel / CSE Student: WONG Ho Nam / SENG Course: UROP1100, Summer

My main task is to implement precise type inference of array indexing in the language MLscript developed by TACO Lab members. This report first introduces the theoretical knowledge I learned by reading the book Types and Programming Languages during the first two months. After that, the report elaborates on my contributions to the language. Lastly, the report describes the possible future work. In short, the report summarizes what I have learned and contributed in this course.

Development on New Compiler for the Artiq System for Quantum Information Experiments (COMP) Supervisor: PARREAUX Lionel / CSE Student: NARAYAN Aadityavardhan / COMP Course: UROP1100, Summer

The purpose of this UROP project was to contribute to the development of NAC3, a compiler for Python programs written for ARTIQ (Advanced Real-Time Infrastructure for Quantum Physics) systems. One of the missing features in NAC3 is the support for static class attributes, which serve as significant tools in programming by enabling the definition of class data members that are independent of any of the class objects. As effective members of the class construct, their implementation will aid the practicality of Python code compiled with NAC3. The aim of this report is to describe the work that has been put into introducing static class attributes into NAC3.

122

School of Engineering Department of Computer Science and Engineering

Acceleration Techniques for Real-time Rendering Supervisor: SANDER Pedro / CSE Student: HU Benran / DSCT Course: UROP3100, Fall

Real-time high-quality rendering in virtual reality is essential for immersive experiences but remains to be a challenging area due to the performance restrictions of untethered head-mounted displays as well as the need for higher framerates and resolutions. On the other hand, perception-based rendering acceleration methods have long been adopted in monoscopic rendering to improve the performance while not sacrificing the perceivable quality. Given that stereo rendering enables us to exploit more properties of the human visual system, it is then a natural choice to boost the performance of VR applications with perception-based rendering methods. In this report, we will examine characteristics of the human visual system that are useful for efficient rendering and elaborate on the design of perception experiments to be performed for verifying and evaluating the feasibility of utilizing these characteristics.

Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: CHOI Sehyun / COMP Course: UROP2100, Fall

Commonsense Knowledge Base (CSKB) Population is a task where it aims to automatically introduce new commonsense entities and relations into CSKB. In our previous work, it was found that incorporation of graph structure into the rich semantics of pre-trained language models led to a promising results. Improving upon this finding, a new graph neural network framework that can effectively capture a larger context of supporting graph information for each candidate knowledge and different pre-trained language model configurations were experimented. The empirical evaluation support the new model framework’s ability to generalize better to unseen examples, while eliciting the existence of spurious bias in the training dataset.

123

School of Engineering Department of Computer Science and Engineering

Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: WANG Yicheng / SENG Course: UROP1100, Summer

This report is about the Attetion techique and the Transformer architecture in Natural Language Processing. To demonstrate Attention, we first introduce a Neural Machine Translation model called seq2seq. And the Attention, as an improvement, is aimed at capture long-term information about the source sentence. Then we go to the Self-Attention technique which reduces the number of unparallelizable operations and acts as a replacement of Recurrent Neural Network. Details and applications of Self-Attention is then illustrated with the Transformer architecture. The building blocks of Transformer, i.e. the encoders and decoders, consist of Self-Attention layers and various additional layers. Transformer is shown to perform better than some old models.

Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: WONG Chi Ho / DSCT Course: UROP1100, Fall

Commonsense knowledge graph (CKG) enhances the performance of many NLP tasks, ranging from question answering to text generation. When it comes to natural language, commonsense components that human possesses are inevitable to provide a more reasonable prediction. However, the difficulties of utilizing Graph Neural Networks (GNN) on the CKG lie in the sparsity of the graph. In the meantime, the higher-order structure between concepts are crucial to generate unobvious concepts. Therefore, in view of these, we propose a pretraining method for GNN on the commonsense knowledge completion task, taking reference from the pretraining strategies of GPT-GNN and UniLMv2.

124

School of Engineering Department of Computer Science and Engineering

Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: XU Baixuan / COSC Course: UROP1100, Summer

Nowadays, Deep learning is getting more and more popular in our daily life. However, how to make the computer understand the commonsense in human daily language remains a tough question in the Natural Language Processing field. To solve this, many computer scientists proposed different heuristic ideas.

As a huge breakthrough in the Natural Language Processing field happened in 2018, the commonsense reasoning field was also significantly influenced by the large pre-trained model brought by Google. This report will mainly illustrate the way of making computers understand commonsense in human language using a large pre-trained language model – Bert by introducing the dataset preparation, question formulation and other related works.

Vision Language Model is another field I delved deeper in during this UROP project period. After CLIP was proposed, the alignment of image and text embedding space have been advanced far a huge step and after that, lots of Vision Language work like VQA, Captioning have been done based on CLIP and its variant. In this report, I’ll briefly introduce the CLIP and the relevant works in Vision Language prompt learning.

Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: JIANG Han / COSC Course: UROP1100, Spring

For a database that contains multi-dimensional data tuples, a user may have his/her unique preference for some data over the others. These preferences are expressed by the users’ utility functions. To determine a user’s favorite tuples in the database, the Strongly Truthful Interactive Regret Minimization method asks questions about a user’s preference on small subsets of the database, and tries to learn his/her utility function as a linear function. We study how to linearize utility functions with specific non-linear forms, in order to adapt these functions into the above method. Specifically, we develop a method which gives all linearizations that meet our purpose. We also present some examples using this method in this report.

125

School of Engineering Department of Computer Science and Engineering

Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: KWAK Byeunggon / COMP Course: UROP1100, Spring

Shortest path index maintenance in dynamic road networks is a problem of designing an oracle that supports the shortest path query efficiently where weights of edges change over time. Dynamic shortest path query is an important operation for various network applications since many real-life networks evolve dynamically. Recently, dynamic contraction hierarchy algorithms achieved improved performance in query processing and index maintenance time compared to other existing algorithms. Particularly, Update Efficient, with randomization techniques, showed favorable theoretical time bounds of 𝑂(log3𝑛) update time with high probability. Nonetheless, shortcut-centric contraction hierarchy outperformed Update Efficient in experimental evaluations, it did not provide theoretical time bounds. In this report, I will analyze the theoretical time bound of the shortcut-centric paradigm using the techniques provided in the analysis of Update Efficient.

Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: LAU Yan Hei / COMP Course: UROP1100, Fall

Session-based recommendation is a task that requires the prediction of a user’s next action based on their previous actions. Conventional methods such as nearest-neighbors and markov chains do not show competent results since they fail to capture all the relational information about items. More advanced methods such as Convolutional Neural Networks and Recurrent Neural Networks show promising results as they are able to use more of the information provided. However, state of the art methods utilize Graph Neural Networks to achieve more context aware systems that outperform all other models. The following report will briefly introduce the structures and purposes of GNNs, and will then provide a summary of three papers on session-based recommendation.

126

School of Engineering Department of Computer Science and Engineering

Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: LU Weiqi / COSC SU Hong / DSCT Course: UROP1100, Fall UROP1100, Fall

In nature, some of the objects have the trend to form colocation patterns, which produces a way to study the co-behaviors of different objects. But only studying the colocation patterns is far not enough. Exploring the causality among the colocation patterns boosts endless value in many areas ranging from ecosystem to business. This report proposes a statistical approach to mine causalities among colocation patterns.

Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: PATUPAT Albert John Lalim / DSCT Course: UROP3100, Fall

Despite several advancements in the field of session-based recommender systems, there is no widely acknowledged assessment framework that employs multiple splits while preventing future information leaking. This report attempted to address this issue by proposing an evaluation protocol with comprehensive instructions and demonstrating its use by assessing three session-based recommendation algorithms based on the k-nearest neighbors approach, namely Item K-Nearest Neighbors (IKNN), Session K-Nearest Neighbors (SKNN), and Sequence and Time Aware Neighborhood (STAN). Results reveal that IKNN is more powerful than previously reported, that SKNN is prone to popularity bias, and that STAN regularly outperforms SKNN. Future work includes applying this proposed evaluation protocol to session-based recommendation algorithms based on recurrent neural networks and graph neural networks.

127

School of Engineering Department of Computer Science and Engineering

Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: SHU Tian / DSCT Course: UROP1100, Summer

The shortest path query is a fundamental problem in spatial networks and is applied in various database and data mining algorithms. As related algorithms often involve frequent queries of the shortest path or distance between two given vertices, a type of data structure called distance oracle is established to store preprocessed information and perform efficient computations of such queries. Literature [1] proposed an efficient distance and path oracle on dynamic road networks named Update Efficient (UE). The oracle employs a randomization technique and supports efficient maintenance of weight updates. This report will briefly introduce some essentials of UE, propose a more efficient update algorithm inspired by literature [2], and provide an analysis of update time complexity with a smaller constant.

[1] Victor Junqiu Wei, Raymond Chi-Wing Wong, and Cheng Long. Architecture-intact oracle for fastest path and time queries on dynamic spatial networks. In Proceedings of the 2020 ACM SIGMOD International Conference on Management of Data, SIGMOD ’20, page 1841-1856, New York, NY, USA, 2020. Association for Computing Machinery. [2] Dian Ouyang, Long Yuan, Lu Qin, Lijun Chang, Ying Zhang, and Xuemin Lin. Efficient shortest path index maintenance on dynamic road networks with theoretical guarantees. Proc. VLDB Endow., 13(5):602-615, jan 2020.

Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: LU Weiqi / COSC SU Hong / DSCT Course: UROP2100, Summer UROP2100, Spring UROP3100, Summer

Co-location patterns and Causality are two well studied fields nowadays. The problem of mining causalities among co-location patterns is an important primitive in knowledge discovery, with wideranging applications from ecosystem to praxiology and epidemiology. However, there are still some limitations in solving the above problem. Many causality mining models lack full utilization of the temporal dimension. And one major challenge of Granger causality test is that it only guarantees the cause precedes the effect in time. To tackle these limitations, this report proposes a novel and scalable model for mining causality from co-location patterns based on Granger causality and designs an algorithm for approximating direct causes from Granger causes based on unique causal information. And extensive experiments are conducted on real-world datasets to evaluate our result and compare it with others.

128

School of Engineering Department of Computer Science and Engineering

Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: WANG Ruida / DSCT Course: UROP1100, Spring

This project marked a starting point for an extensive study project in session-based recommendation systems that eventually targeted some new academic research outcomes and aimed at publishing a paper at some world-top conferences on Machine learning. This semester, the central topic is researching potential filed for research by studying existing pieces (including session-based recommendation system, Distance Oracle on 3D space, Co-location Pattern Mining, Image Retrieval, etc.) and picking the field of session-based recommendation system as the target field for study. The second step in the project is about learning fundamental tools and packages in Machine Learning (including PyTorch, Keras, and Scikit-Learn) together with basic knowledge of the statistic of machine learning. The third step of research in this semester is building some models based on some potential ways of improvements to the current model and testing them with real-world data.

Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: YEUNG Cheuk Long / COMP Course: UROP1100, Fall

Session-based recommendation predicts users’ next actions given a sequence of previous actions in the same session. Session-based recommendation systems using different models have been built in recent years and widely used in major online shopping websites. Research has been done on such recommendation systems at HKUST. This research project aims at providing insights into online shopping patterns using session-based recommendation systems built at HKUST with a web application that simulates online shopping. This report documents the progress of the application.

129

School of Engineering Department of Computer Science and Engineering

Comparative Studies of Programming Languages Supervisor: ZHANG Charles Chuan / CSE Student: TANG Zichen / COMP Course: UROP1000, Summer

Memory safety has been a critical issue that programmers focus on when developing and using the programming language. For programming languages like C/C++ with the support of arbitrary memory access, frequent memory safety bugs come with their outstanding performance. Solving the memory safety bugs while maintaining efficiency and usability is pretty challenging. With the new features of the ownership mechanism and borrow checker, Rust has been a prevalent and efficient programming language for memory safety protection. This report will be mainly about the innovative features in Rust memory-safety issues and the characteristics of those bugs.

Gesture Recognition Using Raspberry Pi Platform Leveraging Machine Learning Technology Supervisor: ZHANG Qian / CSE Student: DING Wenxuan / COMP Course: UROP1100, Fall

Recent gesture recognition techniques are employing acoustic signals to keep track of hand movements and acoustic-based recognition promises to be applied to a wide range of areas, including smart health and smart home. In this project, I designed a few gestures that can satisfy some basic uses of e-guides in shopping malls and developed an acoustic-based gesture recognition system to allow contact-free manipulations of these eguides, which avoids possible virus affection through direct touching. In this recognition system, I used SVM machine learning packets from Python to tackle with the basic classification of gestures and some further analysis into the data collected to allow more advanced commands.

130

School of Engineering Department of Electronic and Computer Engineering

Department of Electronic and Computer Engineering Federated Learning over Wireless Networks

Supervisor: CAO Xuanyu / ECE Student: LIN Yiqiao / CPEG Course: UROP1100, Spring UROP2100, Summer

Traditionally, federated learning aims to train a single global model with several local clients and a server working collaboratively without raw data exchanges. However, this mechanism only develops a single common model for all users, not customized to each client’s data. Previously, we proposed an algorithm named ETCPer-FedAvg, which can greatly reduce the communication overhead while finding the meta-model. This report continues the progress from previous term and gives a more detailed analysis of the proposed algorithm. We evaluate how the performance of the algorithm is affected by the closeness of underlying distribution of user data, triggering threshold and compression operator in nonconvex smooth case.

Projects in Audio Signal Processing - Hearing Loss Simulation and Compensation with Wavelet Transform Supervisor: CHAU Kevin / ECE Student: CHAN Kelvin Cheuk Kwan / SENG MOK Ching Hei / SENG Course: UROP1000, Summer UROP1000, Summer

Hearing compensation is implemented in hearing aids, with most of them using Fourier transform. However, wavelet transform with multiresolution analysis provides greater balance between the time-frequency resolution for the modelling of real-world sounds. Moreover, wavelet transform provides a faster runtime than Fourier transform. To understand more about hearing and hearing loss, hearing loss simulation and compensation were first done in the frequency domain. After that, attempts were made to migrate the simulation and compensation function from the frequency to the wavelet domain. A proposition for a wavelet-based model for hearing loss simulation and compensation, as well as its limitations, were discussed in the paper.

131

School of Engineering Department of Electronic and Computer Engineering

Projects in Audio Signal Processing - A Multi-Band Approach to Improving Speech Intelligibility Using Wavelet-Based Algorithm Supervisor: CHAU Kevin / ECE Student: DU Tianyuan / CPGBM Course: UROP1100, Fall

This research report summarizes some improvements made based on the previous research and conference paper by Yijia Chen et al. from our UROP team, which explored a method for improving the speech intelligibility in the presence of background interference or hearing loss. The method applies a real-time wavelet-based algorithm to separate the speech audio signal into multiple sub-bands, which can then be amplified independently. Since background interference and hearing loss reduce voice intelligibility differently in different frequencies, the multi-band wavelet method allows a more accurate restoration of the original signal, thus resulting in considerable improvements compared to amplifying the whole speech uniformly. The original paper was highly dependent on Google Speech-to-Text recognition as an evaluation means for speech intelligibility. This report supplements further simulations to justify this usage.

Nanostructured Chemical Sensors and Sensor Systems Supervisor: FAN Zhiyong / ECE Student: CAI Yuxiang / ELEC Course: UROP2100, Fall

Following the process of my projects in the summer semester of 2020-2021 academic year, the work participated by me is mostly based on the same principle of AAO (anodic aluminum oxide) gas sensor. The application of gas sensor in more realistic circumstance is focused compared with my previous work.

Although development of AAO gas sensor in my team emphasises on material structure, the design work conducted by me is mostly on PCB layout design, which provides periphery test circuits for severial several projects. This circuits create the interface between sensors and development boards. And the final objective is to transmit these datas collected from sensors to the PC and process the information for display on app, which is complished by my groupmates.

Besides, some ordinary works are also required to be conducted. For instance, bonding and welding the components is necessary for test whether the design can function as expected. Personally, a large variety of Laboratory laboratory rules and flows are learned from many senior students in our group.

132

School of Engineering Department of Electronic and Computer Engineering

Nanostructured Chemical Sensors and Sensor Systems Supervisor: FAN Zhiyong / ECE Student: YANG Zheng / ELEC Course: UROP2100, Fall UROP3100, Spring

CMOS is the dominant technology used to build integrated circuit which is all around our life and perovskites are promising optoelectrical materials which many researchers are working on. Field effect transistors are basic units in CMOS technology and perovskite materials can be used as FET’s channel materials and show similar properties with silicon-based FET. In this report, perovskite field effect transistor’s structure and different perovskite materials used for FET are introduced. And the relationship between insulator dielectric constant and field effect is elaborated. Then hysteresis is introduced and the mechanisms and solutions are talked about. Finally perovskite FET stability and how to measure it are presented.

Robust and Generalized Methods for Medical Image Analysis Supervisor: LI Xiaomeng / ECE Student: CAINE Wilbert / COSC Course: UROP2100, Fall

A Context-aware system is crucial in Operating Room (OR) to ensure situational awareness and promote patient safety. To achieve this objective, the tasks that need to be addressed are dealing with present and future events. There has been massive development on tasks that focus on the present events, such as surgical workflow recognition. However, the high-level future awareness, such as surgical scheduling, is still limited. In one of the previous works, future frames in synthetic surgical videos have been found to be well generated. In this project, we applied two different models to predict the future frames in surgical videos. Future frame prediction can also be utilized in other tasks, which usually requires that future information is available.

133

School of Engineering Department of Electronic and Computer Engineering

Robust and Generalized Methods for Medical Image Analysis Supervisor: LI Xiaomeng / ECE Student: MA Wanqin / ELEC Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

In this summer, we are assigned to the group led by mentor Yao Huifeng. Ma Wanqin focus on preparing Cardiac MRI Analysis Challenge under Respiratory Motion (CMRxMotion), MICCAI 2022. Mu Xihe is assigned to read papers about self-training and consistency learning (eg: pseudo label and mean teacher). To research well, the main goal is to apply what we learned with Professor. Li Xiaomeng during past UROP. After careful selection, we finally decide to use method CACPS to train effective model and participate in the challenge. The method mainly focus on practicing effective segmentation on medical images with the use of pseudo labels. This report will provide challenge information and the details of experiments I did in summer.

Robust and Generalized Methods for Medical Image Analysis Supervisor: LI Xiaomeng / ECE Student: SONG Chen / CPEG Course: UROP1000, Summer

The project reproduces the results of the paper “Not End-to-End: Explore MultiStage Architecture for Online Surgical Phase Recognition” on datasets Cholec80 and M2CAI16. To better understand and complete the project, some related papers and primers are read and learned, also included in this report. This report includes a summary of what I learned. Then there is a summary of the paper, my understanding of the paper and the code. Eventually, it is the experiment and the results on datasets Cholec80 and M2CAI16.

134

School of Engineering Department of Electronic and Computer Engineering

Robust and Generalized Methods for Medical Image Analysis Supervisor: LI Xiaomeng / ECE Student: SUN Haosen / DSCT ZHOU Jialu / DSCT Course: UROP1100, Spring UROP2100, Summer UROP1100, Spring UROP2100, Summer

The sum of the data outside the distribution is the challenge of the machine. This is because most of the learning algorithms so far are based on the assumptions of I.I.D. source or target data. However, under actual conditions, there are usually distribution differences between training sets and test sets. Domain adaptability is a method of solving such problems, but it is necessary to understand the test data. But in many cases, we know nothing about the test data. Therefore, in recent years, many people have begun to consider a more practical research scene called domain generalization. In this article, we have a more comprehensive understanding of the concept of domain summary by analyzing several articles about domain generalization and their respective innovation points.

Robust and Generalized Methods for Medical Image Analysis Supervisor: LI Xiaomeng / ECE Student: YANG Lin / COMP Course: UROP2100, Fall

Image restoration is an elementary tool for numerous low-level visionrelated tasks. It involves generating a high-quality clean image from a given low-quality degraded version. The degradation may occur during the picture capturing, transmitting, and preserving, resulting in different restoration tasks, such as denoising, deraining, dehazing, and deblurring. In this article, the development of image restoration tasks is reviewed, and two representative models will be focused on. One of them is an MLP-based model, which shows stateof- the-art performance on various restoration tasks, and the other one is a transformer-based model, which is all-in-one and degradation-blind. In the end, the future challenges of image restoration will be discussed.

135

School of Engineering Department of Electronic and Computer Engineering

Robust and Generalized Methods for Medical Image Analysis Supervisor: LI Xiaomeng / ECE Student: ZHAO Haihan / CPEG Course: UROP1000, Summer

The current project aims to employ the convolutional neural network ResNet-18 to solve the classification task of the CIFAR-10 dataset, a collection of 60000 32*32 RGB images commonly used to train and test typical neural networks, including KNN, linear, and convolutional neural networks. The network was built with PyTorch and executed on Google Colab. This report reviews relevant literature and elaborates on a sample code, including a data loader, a ResNet model, and a train function. The final accuracy is roughly 80%. With the addition of information argumentation and module improvement, it can achieve more satisfactory results and be applied in long-tail recognition tasks.

Learning with Limited Anotated Data for Medical Image Diagnosis Supervisor: LI Xiaomeng / ECE Student: GUO Jiarong / ELEC Course: UROP2100, Fall UROP3100, Spring UROP4100, Summer

Labeling relatively large datasets always take a significant amount of time, especially in medical images. Semi- Supervised Learning (SSL) improves results by utilizing both labeled and unlabeled data. Previous SSL methods assume the labeled and unlabeled data are in a closed set. However, open-world unlabeled data can contain classes of out-of-set data which transfer negative results in unlabeled learning. It is Open-set Semi-Supervised Learning (OSSL). In this paper, we address an exciting but more challenging setting, OSSL and multilabel classification, which is a more realistic setting. We propose a framework for novel class detection and multi-label pseudo-label generation to solve the multi-label semi-supervised learning and open-set problem. We capitalize on the knowledge obtained from labeled data and distribution in feature space. By conducting several experiments on multi-label datasets with retinal and natural images, our methods can significantly improve the model's ability to perform during the evaluation stages. Our code will be released on GitHub.

136

School of Engineering Department of Electronic and Computer Engineering

Artificial Intelligence Methods for Medical Videos Supervisor: LI Xiaomeng / ECE Student: HUANG Junkai / COSC Course: UROP1100, Fall

Phase recognition of surgical video is of particular interest to computer assisted surgery systems. Current fully supervised methods require the training data to be manually labeled by human experts, which is a tedious and time-consuming work. In this project, we aim to develop a surgical video phase recognition method that can be trained under a weakly-supervised setting, in which only the order of phases appears in the video is provided during training. We have conducted experiments on state-of-the-art video phase segmentation pipelines and methods including MSTCN, Timestamp supervision, and Cross Pseudo Supervision. All the experiments are performed using Cholec80 dataset. Then we analyze the results and look for possible chances for improvements.

Artificial Intelligence Methods for Medical Videos Supervisor: LI Xiaomeng / ECE Student: LIU Yunfei / CPEG Course: UROP1100, Spring UROP2100, Summer

An adversarial attack is a technique to create adversarial models. Thus, an adversarial model is a contribution to an AI model that is deliberately intended to make a model commit an error in its expectations notwithstanding looking like a substantial contribution to a human. Thusly, to stay away from such bothers, adversarial training is especially significant, which is a savage power regulated learning technique where however many adversarial models as could be expected under the circumstances are taken care of into the model and expressly marked as compromising. This is a similar methodology the normal antivirus programming utilized on PCs utilizes, with different updates consistently. In this article, I will present a report about adversarial training that I have perused and comprehend in UROP2100.

137

School of Engineering Department of Electronic and Computer Engineering

Artificial Intelligence Methods for Medical Videos Supervisor: LI Xiaomeng / ECE Student: LIU Ziming / CS/COMP Course: UROP1100, Fall UROP2100, Spring

In this UROP 2100 project, I was assigned tasks on surgical video analysis, focusing on future frame prediction. My work basically consists of reproducing the paper “Surgical Workflow Anticipation Using Instrument Interaction” and finding ideas on improving the baseline by reading other papers. I will present them in two parts in the following sections.

Artificial Intelligence Methods for Medical Videos Supervisor: LI Xiaomeng / ECE Student: ZHANG Yuyao / DSCT Course: UROP1100, Spring

This semester, since my background knowledge of deep learning was not enough, I just read several papers and learned the basic knowledge. So in this report, I will write a summary of some of the papers and some models.

Deep Learning for Whole Heart Segmentation and Disease Diagnosis Supervisor: LI Xiaomeng / ECE Student: KUSUMAWARDHANI Adelia Savitri / SENG Course: UROP1000, Summer

Advance medical imaging modalities are utilized to provide information regarding the anatomical structures of the heart, occupying an essential role in the prevention and medication of cardiac diseases. Despite the general outstanding accuracy of Deep Neural Networks (DNNs), medical segmentation data suffer from a shortage of data annotation and scarcity in their pixel categories. We believe combining the semi-supervised method from a previously published paper and implementing the tau norm classifier would help achieve better accuracy and bridge the gap in real-world data collection scenarios. In this paper, we will not only evaluate the performance of the tau norm classifier through comparison but also observe the accuracy attained from different tau values.

138

School of Engineering Department of Electronic and Computer Engineering

Deep Learning for Multi-class Retinal Disease Classification in Real-World Setting Supervisor: LI Xiaomeng / ECE Student: LIU Zichen / DSCT Course: UROP1100, Spring

This semester, I participated in UROP under Professor LI, Xiaomeng's supervision. My work includes reading relevant papers, implementing and reappearing recent papers' experiment results, and conducting experiments on novel ideas. Specifically, I read more than 30 latest works about medical image and computer vision. I reappeared the result of the paper “A Fourier-based Framework for Domain Generalization” and “RSCFed: Random Sampling Consensus Federated Semisupervised Learning”. I also experimented to verify that MMD loss can improve the result of federated semi-supervised learning classification. In this final report, I will illustrate my works in detail.

Deep Learning for Multi-class Retinal Disease Classification in Real-World Setting Supervisor: LI Xiaomeng / ECE Student: ZHANG Yuyao / DSCT Course: UROP2100, Summer

I read some papers during this summer UROP program and participated in a challenge. Therefore, in the first part of the report, I will summarize some papers, and in the second part, I will discuss the challenge.

Federated Learning with Medical Images Supervisor: LI Xiaomeng / ECE Student: MU Xihe / CPEG Course: UROP1000, Summer

In this summer, we are assigned to the group led by mentor Yao Huifeng. Ma Wanqin focus on preparing Cardiac MRI Analysis Challenge under Respiratory Motion (CMRxMotion), MICCAI 2022. Mu Xihe is assigned to read papers about selftraining and consistency learning (eg: pseudo label and mean teacher). To research well, the main goal is to apply what we learned with Professor. Li Xiaomeng during past UROP. After careful selection, we finally decide to use method CACPS to train effective model and participate in the challenge. The method mainly focus on practicing effective segmentation on medical images with the use of pseudo labels. This report will provide challenge information and the details of experiments I did in summer.

139

School of Engineering Department of Electronic and Computer Engineering

Design of Radio-frequency Integrated Circuits Supervisor: LUONG Howard Cam / ECE Student: LI Cunxu / ELEC Course: UROP1000, Summer

In the DESIGN OF RADIO-FREQUENCY INTEGRATED CIRCUITS project, l followed the professor’s course to get a sense of what's involved. According to the knowledge I have acquired, l choose to mainly focus on the Inductors and Varactors. I understand how they are used in the real world and learn how to design them, how to solve the problem in designing them and their limitations and shortcomings. I got a further look at the circuit design. Because of my limited knowledge, I can't go further this summer. But with my further study in the following semester, I am willing to continue working on the part I currently can’t understand.

Design of Radio-frequency Integrated Circuits Supervisor: LUONG Howard Cam / ECE Student: XING Hexiang / ELEC Course: UROP1100, Spring

My urop project is the design of radio-frequency integrated circuit. The increasing demands of highly efficient communication push the operation band of the receiver to mm-wave frequency. Wideband LNA is the key block in the mm-wave receiver. Thus, I mainly focus on the design techniques of wideband low noise amplifier (LNA) design. In the first part, some RF basics for LNA design are summarized such as the noise figure, linearity and sensitivity. I give a detailed analysis of the basic topologies for wideband LNA in the second part. In the third part, several advanced LNA design techniques are introduced, such as noise-canceling, current reuse, gm-boosting, etc. Finally, in the fifth part, introduction of my wideband LNA design is given.

140

School of Engineering Department of Electronic and Computer Engineering

An Optofluidic Biological Cell Stretcher Supervisor: POON Andrew Wing On / ECE Student: ZHU Guanhua / CPEG Course: UROP2100, Fall

During the last project from UROP1100, we have illustrated the setup of an opto-fluidic single-beam biological cell stretcher, the theory behind the major building blocks of the setup, built the cell stretcher and demonstrated the actual applications of the device. This project focuses on the progress of the last project, including improving the imaging system of the biological cell stretcher, method of processing the images obtained from the imaging system, cell recognition and data extraction from the images using the vision toolbox of MATLAB and the exploration of future works by analyzing the state-of-the-art optical tweezers from other institutions.

Circuit Emulation of Bio-Inspired Dynamic and Topological Quantum Systems Supervisor: SHAO Qiming / ECE Student: FU Penglin / SENG Course: UROP1000, Summer

In this work, I use electronic components like memristor, MOSFET, op-amp, etc., to simulate a circuit that can perform weighted addition operations. Crossbar structures are used in my work for their space-saving and energy efficiency. Both positive and negative and 0 values of weights are achieved in this work. Also, an approximate activation function can be realized in this work. The training process is not done on this device. Weights and biases are pre-calculated and tuned into the device before usage. By sending input and getting the output of this device, a result the same as a fully connected layer is generated.

Circuit Emulation of Bio-Inspired Dynamic and Topological Quantum Systems Supervisor: SHAO Qiming / ECE Student: LIU Yifan / SENG Course: UROP1000, Summer

The project should have been designing a practical Operational Amplifier (Op Amp) at first. However, I was not able to figure out everything from scratch due to numbers of problems occurred in the progress. But fortunately, I find the final design of my Op Amp is quite inspiring and I think it could be helpful for beginners to understand how an Op Amp works. Besides, after reading some text books, I think there lack of some simple methods for designing only an Op Amp with MOSFET. Therefore, in this report, although the Op Amp is not practical, it can guide the beginners to design their own Op Amp in a much shorter time. There will be sections around ‘Principle of MOSFET, ‘Optimal Operational Amplifier’, ‘Topology of Op Amp’, ‘Differential Amplifier’, ‘CommonSource Amplifier’, ‘Parallel Resistance’, ‘Class AB follower (In Progress)’ and ‘Verification’.

141

School of Engineering Department of Electronic and Computer Engineering

Circuit Emulation of Bio-Inspired Dynamic and Topological Quantum Systems Supervisor: SHAO Qiming / ECE Student: ZHENG Yueyan / CPEG Course: UROP1100, Fall UROP2100, Spring

In recent years, non-volatile memory has gained more attention. Compared to volatile memories SRAM, DRAM, non-volatile memories tend to be large, slow to read and write, and consume more energy. Magnetoresistive Random Access Memory (MRAM), on the other hand, offers improvements in all of these areas. As a memory designed to exploit the spin characteristics of electrons, MRAM can use the spin transfer torque (STT) effect to transform stages by applying current, while the voltage-controlled magnetic anisotropy (VCMA) effect can be used to write by applying voltage. This STT-assisted VCMA MRAM is characterized by low power consumption, high switching speed and high density. In terms of stability, it is worth investigating further.

Circuit Emulation of Bio-Inspired Dynamic and Topological Quantum Systems Supervisor: SHAO Qiming / ECE Student: ZHENG Wenhan / CPEG Course: UROP2100, Fall

This semester, the UROP project involves using multiple instruments to measure unknown devices to get the properties for further analysis. Luckily, there is a certain part designed for TLM measurement to get contact resistance between metal and semiconductor and unit resistance of semiconductor material. The raw resistance data are obtained by remotely-controlled K2636B using python program. The dimensions of test devices are measured by microscope. Then the raw data are processed by using linear approximation. The final approximated data reviled the resistive characteristics of the semiconductor part. However, the measured data of contact resistance contradicts the estimated value from linear approximation. This is partially caused by different metal-semiconductor junction behaviors.

Compact Models for Circuit Design Supervisor: SHAO Qiming / ECE Student: TAN Tianshu / SENG Course: UROP1000, Summer

The purpose of this thesis is to report the design procedures of a 2-stage operational amplifier with smic180 nm under the requirements of the SSCS student circuit design contest 2022. The amplifier is constructed by 7 MOSFETs, 2 resistors, 1 capacitor, and 1 ideal voltage source. It has attained a 38.42KHz bandwidth, a 103.1° phase margin, and a 37dB gain.

142

School of Engineering Department of Electronic and Computer Engineering

Compact Models for Circuit Design Supervisor: SHAO Qiming / ECE Student: XING Hexiang / ELEC Course: UROP1100, Fall

As CMOS technology scales down, the power dissipation due to the leakage current becomes a serious concern, and spintronic devices are promising candidates for solving this issue. Spintronic devices have their own advantages such as high endurance, nonvolatile characteristic, low power dissipation, and good scalability. Among various spintronic devices, magnetic tunnel junction (MTJ), as the essential unit for magnetic random-access memory (MRAM), can be applied for various fields such as memory, in-memory computing, data security, and stochastic computing. Currently, spintransfer-torque MTJ (STT-MTJ) is widely applied. However, compared with STT-MTJ, voltage-control MTJ (VC-MTJ) can achieve higher density, lower power consumption, and faster speed. In this report, I summarize some basic applications of VC-MTJ in different devices such as magnetoelectric random-access memory, true random number generator, analogto- digital/stochastic converter, spintronic programmable logic, temperature sensor, in-memory computing, etc.

Compact Models for Circuit Design Supervisor: SHAO Qiming / ECE Student: XU Zimo / CPEG Course: UROP1100, Fall

DNN+Neurosim is a kind of integrated Framework that created by c++ language. It uses python wrapper to connect the neurosim and a machine-learning platform, Pytorch, to provide a mapping from algorithm to hardware that could be used to evaluate the chip training in area, energy efficiency and throughput.

Deep Learning for Magnetic Domain Image Denoising and Super-Resolution Supervisor: SHAO Qiming / ECE Student: DUAN Qinkai / SENG Course: UROP1100, Summer

I focused on learning during this project: autoencoder, t-sne, LSTM, Restricted Boltzmann Machine, and mumax3 tutorials. They are about graphic classification, time list prediction, and magnetic simulations. In many datasets, the dimension of the data is too big to visualize. So we need t-sne method to reduce the dimension to 2 or 3 while keeping the distribution. LSTM is an extension for RNN. It is used for predicting the next time step action in a time interval. It’s widely used in word and text recognition. Mumax3 can simulate how the magnetic material change in atom size based on equations in Micromagnetics.

143

School of Engineering Department of Electronic and Computer Engineering

Deep Learning for Magnetic Domain Image Denoising and Super-Resolution Supervisor: SHAO Qiming / ECE Student: PENG Yiyan / SENG Course: UROP1000, Summer

This report introduces what I learned this summer semester and the Handwritten Digits Recognition algorithm. The new knowledge that I learned focuses on cost function, gradient descent, linear and logistic regression, and Regularization. The algorithm is the application of fundamental machine learning knowledge. The model’s dataset is provided by MNIST, and the network type is the fully connected neuron network. Some improvements like Regularization are also presented in the end.

Embedded Systems for AI Hardware Supervisor: SHAO Qiming / ECE Student: LAU Wai Yin / ISD Course: UROP1000, Summer

This report is targeted at a proposal to produce a main module for a high-speed switching controller that is suitable for AI algorithm. It have taken references from a published paper(https://aip.scitation.org/doi/abs/10.1063/5.0058786), a FPGA board provided 1ns time resolution and maximum bandwidth up to 125Mhz, included 14 digital output channels and allowed data sequences for diverse commands. Therefore, this proposal is targeted to propose a similar device with literature reviews and potential electrical components selection mentioned. The whole design will be built according to the open-source code provided from the reference paper as a good starting point. Additional circuit will be considered to suit our user scenario.

Embedded Systems for AI Hardware Supervisor: SHAO Qiming / ECE Student: SHI Yiqi / ELEC Course: UROP1100, Summer

This is a study report which reflects what I have done in Summer 2022 UROP1100 program. My UROP topic is ‘Embedded Systems for AI Hardware’. My main job in this semester is reading and learning, which means this is neither a project report nor progress report. Instead, it is a study report. I have come over with four weeks of selfstudy and three weeks of thesis reading. In the following report, I will list them out one by one.

144

School of Engineering Department of Electronic and Computer Engineering

Dirt Classification Scheme on a built-up Robot Platform Supervisor: SHI Ling / ECE Student: CHANG Mincheol / ELEC Course: UROP1100, Fall

Developing a well-performed dirt classification scheme is a fundamental task for cleaning robots nowadays. The technique of dirt detection allows locomotion robots to identify and fulfill the task of cleaning in the first place, and avoid hard-to-clean objects. This report proposes a method to build a dirt dataset and dirt classification scheme using a neural network model for the embedded system of a locomotion robot. The report illustrates several detection model candidates such as Tiny YOLOv2, Mask R-CNN, and simple MobileNet, which can be applied to embedded systems with the robot's KPU chip.

Multi-agent System Control Supervisor: SHI Ling / ECE Student: LIU Hanmo / CPEG Course: UROP1100, Spring

During the actual operation of the unmanned vehicle, the laser radar input is used to determine the location of unknown obstacles, so as to further calculate obstacle avoidance and path planning. The lidar input is the feedback length of the target point and the angle between the target point and the car. A current simulation environment is a number of cylindrical target obstacles with a radius of 0.05m. The trolley needs to identify and record these obstacles during the traveling process for subsequent elfslam calculation.

Multi-agent System Control Supervisor: SHI Ling / ECE Student: WANG Peiqi / ELEC Course: UROP1100, Fall UROP2100, Spring

The last decade has seen a rapid development of the Automated Guided Vehicle (AGV). AGVs can help improve the throughput of warehouses while reducing the labor cost. In this report, we will be focusing on some of the common methods of indoor AGV path planning. The survey will serve as a foundation for further research in mobile robot navigation and help the researchers build more advanced vehicles for complex tasks.

145

School of Engineering Department of Electronic and Computer Engineering 146 Multi-agent System Control Supervisor: SHI Ling / ECE Student: WANG Yucheng / COSC Course: UROP1100, Spring

With the rapid development of modern technology, the robotics industry is bursting with vigor. Various types of robots appear in our lives, and they serve all aspects from diet to living, from production to assembly, from land to sea and space. Under the guidance of Professor SHI Ling, we designed a cruise robot this semester, which can carry heavy objects for automatic navigation and obstacle avoidance. We believe that this robot will have many application scenarios at HKUST, from the delivery of books in the library to the transportation of items between laboratories. In the future, with further design and installation of robotic arms, 'Ling Dynamic' will usher in more tasks and challenges.

Multi-agent System Control Supervisor: SHI Ling / ECE Student: YUN Hojin / CPEG Course: UROP1000, Summer

A multi-agent system is an emerging field in robotics. A problem that could not be solved with a single robot can be solved with multiple robots, such as collaborative object carrying, and search and rescue. Under the supervision of Professor Shi ling, this project’s goal is to design a multi-agent system composed of two robots finding their pathway to a target box’s location. Upon reaching the destination, two robots will coordinate to move the box and return it to a target location. The project could be further developed to carry more complex shaped objects or carry heavier objects with more robots, and could be used as This report will discuss the current progress of the project.

146

School of Engineering Department of Electronic and Computer Engineering

Multi-agent System Control Supervisor: SHI Ling / ECE Student: ZHAO Xiaoqi / ELEC Course: UROP1100, Fall UROP2100, Spring

Multi-agent systems (MASs) have received increasing attention from scholars in different disciplines. A platform is set up to test algorithms on multi-agent robot systems, such as formation control and cargo delivery. The platform uses Vicon motion capture system for the positioning of multiple differential wheeled robots. We study the multi-robot collaboration applications on the platform with the configurations similar to an industrial warehouse environment. The robots can now move a cargo from one place to another collaboratively by modeling the formation similar to the kinematics of a differential wheeled robot. The work utilizing centralized control method lays a foundation for future research on decentralized multi-robot collaboration control as well as works on SLAM-based positioning method.

RFID-based Indoor Localization for UGV Supervisor: SHI Ling / ECE Student: KUI Ka Yu / SENG LEUNG Ka Chun / SENG LI Kin Fung / COGBM Course: UROP1000, Summer UROP1000, Summer UROP1000, Summer

RFID applications are everywhere. From small-scale applications such as door access control, to large-scale applications such as warehouse storage. RFID plays a vital role in identification and localization and is widely adopted. Because of RFID’s cheap, easy to program property, RFID can also be a powerful educational tool for students to learn the basics of coding and algorithm interactively. This report describes our work in the Undergraduate Research Opportunity Program in the 2022 Summer. First, explain the rationale of RFID, then how we can leverage the properties of RFID to make a robot and RFID carpet for STEM education and our future works.

147

School of Engineering Department of Electronic and Computer Engineering

Neural Coding for Motor Brain Machine Interfaces Supervisor: WANG Yiwen / ECE Student: WANG Ziyi / DSCT Course: UROP1100, Spring

The project “Neural Coding for Motor Brain Machine Interfaces” aims to build neural coding models for motor brain machine interfaces. Neural coding consists of 2 related parts, encoding and decoding. Encoding means to predict neural codes from kinematics data, and decoding means the reverse, to predict kinematics data from neural codes. In this semester, the project focuses on the encoding part, and is still in progress. This report shows the algorithms and theorems for encoding, the process of simulation to verify the feasibility of the algorithm, and the manipulation of parameters to understand the influence of every parameter on the outcomes.

Development of Network Communication System for Locomotion Robots Supervisor: WONG Man / ECE Student: WANG Haoyi / CPEG Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

The aim of the project is to use UWB positioning technology to design a wireless system for robot localization. Our research proposed a new improved method using unified modules without the differentiation as base stations and tags. In this phase of the project, we thoroughly explored and modified the program embedded in the UWB modules, so that not only can we acquire the distance between any pair of modules, but also, we can change the output from the serial port to get any kind of information we want. We also did multiple experiments to get a large amount of data to estimate the performance of the modules in different conditions.

Design and Characterization of a Tactile Sensor Array Supervisor: WONG Man / ECE Student: CHEN Sida / ISDN Course: UROP1100, Fall

The research topic I’m helping with is called tactile sensor array. It is a kind of sensor array that can detect the force applied on top of it. It was made with the technology called Thin-Film Transistor (TFT) and as a result, its circuit only contains NMOS instead of CMOS. I was responsible of design the testing environment and conduct characterization of the manufactured die. There are two things to test in this project: sensor array and the ADC circuit. I would elaborate on what I have done, what are the difficulties, and what I have learned in this project.

148

School of Engineering Department of Electronic and Computer Engineering

Nano/Microfabrication Technologies for Microelectronic and Photonics Supervisor: YANG Yansong / ECE Student: HO Tsz Fung / CPEG Course: UROP1100, Spring

This report summarizes the effort paid to receive training at the Nanosystem Fabrication Facility in the University, as well as to gain the qualifications required for conducting microfabrication laboratory works, under the social restrictions imposed during the pandemic. In this project, the safety, chemical and contamination control trainings, and their relevant examinations were scheduled and received to obtain the user approval of the Facility. Trials using different microfabrication techniques had also been put for implementing the process flow for patterning aluminum electrodes on the irregular shaped silicon wafer, while contact photolithography, E-beam lithography, sputtering, e-beam evaporation, oxidation, wet etching, dry etching, and lift-off technique was investigated during the process.

High-performance CMOS Transimpedance Amplifier Design for Short-Range Optical Communications Supervisor: YUE Chik Patrick / ECE Student: NUGRAHA Ferris Prima / CPEG Course: UROP1100, Fall

This research project aims to investigate the electronic components involved in antenna systems that could be applied for Wireless Power Transfer (WPT) and Positioning. WPT is becoming ubiquitous, especially in the IoT for low-powered logistics tracking or more challenging power transmission of wireless robots. In this project, we focus more on the patch or microstrip antenna arrays responsible as the power transmitter at the power beacon through electromagnetic radiation in the frequency regime of 2.4GHz and 5.8GHz, beamformed to the receivers located in specific location. Several characteristics of the antenna structure studied include the far-field radiation pattern, directivity, efficiency, and the S-parameters. Selected antenna structures from papers could be simulated with state-of-the-art software like CST Studio Suite 2020.

149

School of Engineering Department of Electronic and Computer Engineering

Design of a Wireless Data and Power Transmission System for IoT Devices Charging and Communication Supervisor: YUE Chik Patrick / ECE Student: KANG Tianqu / EEGBM Course: UROP1100, Fall

The goal of the project is to build one MWPT system that can charge the devices wirelessly while transmitting data between the charger and receiver as well. Bluetooth is used for the connection and charging, as it is easy to implement, and has high indoor localization accuracy. To perform wireless charging, the first step is to connect and locate the receiver. This report would focus on Bluetooth Indoor localization. The general wireless localization method is firstly analyzed, then the accuracy of the RSSI based localization method would be evaluated.

Video Compression via Deep Learning for Resource-constrained Edge AI Systems Supervisor: ZHANG Jun / ECE Student: ZHANG Yutian / CPEG Course: UROP1100, Spring UROP2100, Summer

The wide deployment of audiovisual devices in recent years has placed increasing demand for data storage and transmission. Meanwhile, the development of artificial intelligence and deep learning technologies has provided promising tools for effective data compression, saving storage cost and reducing transmission latency. Neural network-based methods can greatly improve the compression efficiency and are bringing a paradigm shift to various data compression problems. This project will focus on low-complexity deep neural networks for video compression, considering resource-constrained edge devices.

150

School of Engineering Department of Industrial Engineering and Decision Analytics

Department of Industrial Engineering and Decision Analytics Remote Vital Signs Supervisor: SO Richard Hau Yue / IEDA Student: CHENG Chun Hong / COMP Course: UROP1100, Fall UROP2100, Spring

Based on the findings and research of previous UROP 1100E project, we dived deeper and analysed different methods for remote blood oxygen saturation (SpO2) measurement. During this UROP 2100F project, we have proposed a deep learning method for remote SpO2 measurement with the use of a spatial-temporal representation ― spatial-temporal map (STMap). We have evaluated this deep learning method with a largescale multi-modal public benchmark dataset, VIPL-HR. Our method outperforms existing traditional remote SpO2 measuring methods and it is also robust in different situations. This shows that our proposed method has a great potential to be applied for healthcare services.

Data Analysis in Fintech Supervisor: ZHANG Jiheng / IEDA Student: LI Hao / DA Course: UROP1100, Fall UROP2100, Spring

The Long-Shot Term Memory (LSTM) networks are used to predict the cryptocurrency prices. Our results show that LSTM gives good predictions for the time series in our study here, and makes much better predictions over the classical methods used in our previous study. Some of the technical issues and future directions are mentioned and discussed at the end of the report.

Data Analysis in Fintech Supervisor: ZHANG Jiheng / IEDA Student: LIU Yuzhi / MATH Course: UROP1100, Fall

The world has seen an explosive development of information technology during past decades. Following this trend, FinTech as well as algorithmic trading has rapidly flourished. The report gives a brief overview of the different cryptocurrencies available today and describes an implementation using a modified trading algorithm on a given database.

151

School of Engineering Department of Industrial Engineering and Decision Analytics

Data Analysis in Fintech Supervisor: ZHANG Jiheng / IEDA Student: WANG Xiaopeng / DSCT Course: UROP1100, Fall

In recent years, cryptocurrency, a collection of binary data which is designed to work as a medium of exchange, has become more important in the financial market. And Binance is a cryptocurrency exchange which is currently the largest exchange in the world in terms of daily trading volume of cryptocurrencies. With the increasing importance of cryptocurrency, it is meaningful for us to look into the change of various cryptocurrencies and develop some corresponding trading strategies. In this project, my goal is to analyze the data distribution first and then develop a practical trading strategy by Python (pandas) based on the database of real financial transaction data (Bitcoins) from exchanges (Binance), which is received by a given IT infrastructure.

Data Analysis in Fintech Supervisor: ZHANG Jiheng / IEDA Student: ZHAO Yanrui / DA Course: UROP1100, Fall

Cryptocurrencies are becoming valuable and demanding. Reviewing current financial industry, where are not many derivative products relating to cryptocurrencies. This research tries to implement a trading strategy for stocks and futures into bitcoin and analyse the profitability. The raw data of trade is summarised and processed into a form that is usable for analysis. Several graphs are drawn using module Bokeh in Python for data visualisation. The performance test is carried out after discovering the trends from the visualised data. Performance results are finally drawn using Bokeh for a clear illustration of the results. It is found that the strategy based on moving average made considerable returns, and the more frequent the actions take the higher overall growth outcome.

152

School of Engineering Division of Integrative Systems and Design

Division of Integrative Systems and Design Mixed Reality Blackboard Supervisor: BRAUD Tristan Camille / ISD Student: FISILO William Arvin / SENG Course: UROP1100, Summer

The aim of the research is to have a more effective learning experience in classroom. Research by Brown et al (June 2016), shows how VR’s effectiveness compares to the traditional classroom models. Thus, in this project, Virtual Reality (VR) is used as the fundamental, to create augmented reality (AR), aid remote learning, and allow the use of the third dimension, comparative to the traditional classroom experience. The increase in VR usage, and how it is a low-cost tool, is the reason for the growing popularity of VR/AR. In this project, users are able to draw/write using VR controller through a virtual space.

Synthetic Tactile Sensing for Prosthetic Fingers Supervisor: LEUNG Suk Wai Winnie / ISD Student: LIU Hanmo / CPEG Course: UROP2100, Fall

Detecting the state of hand activity and adjusting the positive and negative feedback has a positive effect on the rehabilitation training of disabled patients. By adding different sensors, the artificial fingers may take the different environmental signals as input such as pressure, fracture, temperature, etc. Sensing the signals is the first step for prosthetic fingers. This research focuses on how to construct one appropriate circuit which can apply to specific sensors to fulfill the requirement of the gloves.

Finger Joints Design of 3D-printed Exo-Prosthetics Supervisor: LEUNG Suk Wai Winnie / ISD Student: KOK Yung Kang / MAE Course: UROP1100, Spring

Exo-prosthesis hand is a replacement for a human hand to help disable to adapt to normal human live. In contrary to the more popular pulley system design in 3D printed exo-prosthesis hand, this project research in the design of linkage driven finger to provide a design which might possibly be used to compare with the pulley system in terms of reliability and user experience. The design shown in this paper still consists of flaws in terms of rigidity and compactness. Further improvement and testing have to be done to completely evaluate the possibility of using linkage system in a low-cost 3D printed protheses hand.

153

School of Engineering Division of Integrative Systems and Design

HKUST in VR Supervisor: YEUNG Sai Kit / ISD Student: LAI Kam Wai / ISDN TSE Yui Kai / ISDN Course: UROP2100, Fall UROP2100, Fall

This project requires the use of Unity and VR equipment to create a virtual environment for laser cutter training. Tutorials on Unity VR are taken to familiarize with the application of the software. Training for manufacturing tools used to be tedious and time-consuming, since one-to-one training is usually required for students to be familiar with the operation, which takes technicians so much time and effort to repeatedly teach individual students. Hence, training using virtual reality might be a way to increase the efficiency of training since students can be trained without technicians and it does not require multiple physical devices for training. After building the 3D models for a variety of the equipment in the summer before, this semester, we designed our user survey and conducted various user tests to test out the effect of the prototype.

154

School of Engineering Department of Mechanical and Aerospace Engineering

Department of Mechanical and Aerospace Engineering Electrochemical Organic Synthesis Supervisor: CHEN Qing / MAE Student: LAM Yan Yi / CHEM Course: UROP1100, Spring

Organic synthesis is essential as it helps develop drugs and provides a more efficient way to synthesize drugs. However, the traditional method of organic synthesis will produce toxic products, byproducts, side products, and chemical waste. This leads to environmental problems. Besides, more severe natural disasters and global warming raise concerns about environmental protection.1 So, with the development of chemistry, green chemistry is getting more critical. Toxic reactants and catalysts are replaced by greener chemicals or avoided by developing alternative pathways. The electrochemical approach is one of the green methods in organic synthesis. The byproducts of electrochemical organic synthesis are mainly hydrogen and oxygen, which are caused by water splitting. In the development of electrochemical organic synthesis, a suitable choice of the electrode is important, which will directly affect the efficiency and selectivity of the target products.

Polymer Na Composite Batteries Stabilizing Mn-based Positive Electrodes Supervisor: CIUCCI Francesco / MAE Student: WANG Ruoyi / CHEM Course: UROP1100, Fall

Lithium-ion batteries (LIBs) have been considered the most promising and successful rechargeable battery since 1991. They have a wide range of industrial applications and possesses several advantages like high voltage, high energy densities and low self-discharge rates. However, in face of mass marketization of gridlevel energy storage systems which demand a large amount of materials, LIBs commercialization has been hindered by lithium ions’ low earth abundance, limiting potential in energy density (estimated to reach its intrinsic maximum within a few years) and thermal instability due to volatile, toxic, flammable organic solvents typically used to increase the ionic conductivity of LIBs. Therefore, exploring possible alternatives to LIBs, this study investigated and proposed polymer sodium composite battery as a potential substitute. We believe this research has contributed to the literature of sodium-ion battery (SIB) as well as dual-salt polymer electrolyte (DSPE). Limitations and possible improvements in experimental design shall be discussed.

155

School of Engineering Department of Mechanical and Aerospace Engineering

Development of Protonic Reversible Fuel Cells Supervisor: CIUCCI Francesco / MAE Student: WANG Heumil / MECH Course: UROP1000, Summer

Proton conducting solid-oxide fuel cell (SOFC) is a modification of ordinary SOFC, characterized by flexible fuel compatibility, non-precious-metal catalyst, and wider range of operating temperature. This report primarily focuses on material synthesis of cathode and electrolyte, X-ray Diffraction (XRD), and symmetric cell testing. Different materials have been prepared as potential cathode candidate, followed up by XRD. Having both faces sprayed with cathode on electrolyte palette, symmetric cell serves solely to test the performance of cathode. Upon obtaining satisfactory performance of cathode via symmetric cell testing, one can subsequently conduct studies on creating and testing performance of fuel cell as well as making it more efficient.

Multiphase Simulations with CFD Supervisor: FU Lin / MAE Student: PANG Ho Kiu Holdee / MAE Course: UROP1100, Fall

Simulations on multi-phase flows such as shock-bubble/droplet interaction and cavitation phenomenon are important in solving modern engineering problems. Such flow phenomena are challenging to deal with using traditional CFD techniques, especially with processes such as droplet generation and moving contact line dynamics. This project aimed to explore the potencies of alternative CFD framework such as the MFC framework, which are particularly useful for dealing with multi-phase and multi-component flow. The software structure, algorithms used, and equations were studied, and the flow code was compiled under various settings of typical flow examples. The data collected were then analysed and used for visualization after the simulation.

156

School of Engineering Department of Mechanical and Aerospace Engineering

High-order Numerical Method for Predicting Flows over Aircrafts Supervisor: FU Lin / MAE Student: TSE Ming Hang Henry / AE Course: UROP1100, Spring

RAE2822 airfoil is a transonic airfoil designed to reduce the effect of shock waves. The report covered extensively steps to produce an analysis of the RAE2822 airfoil and a comparison with experimental data was made. Open-source programs OpenFOAM and ParaView were used for the study. From setting up the software to creating an analysis case, it was not as easy as following the instructions, there were commands to be understood and modifications were made. Using a tutorial case as a reference and online resources, RAE2822 was created digitally for analysis. Although the simulation result might not be identical to the experimental result, potential reasons were suggested.

High-order Numerical Method for Predicting Flows over Aircrafts Supervisor: FU Lin / MAE Student: ZHANG Rushan / AE Course: UROP1100, Fall

This project consists of 2 stages. The first stage is to use a computer program written by Doctor Cheng Cheng to analyse NACA 0012 airfoil. With the results from the program, the lift curve and the drag polar were generated, and were used to compare with reference data collected under similar parameter settings. The second stage is to draw the mesh for a three-element high-lift airfoil, McDonnell Douglas 30P30N airfoil. In this stage, two types of mesh were generated. The first one is a hybrid mesh, which consists of both rectangular regions and quadrilateral regions. The second one is a structured mesh, which consists only of quadrilateral regions.

Extended Ortho-K Lens for Myopics Supervisor: LAM David Chuen Chun / MAE Student: SEO Minji / AE Course: UROP2100, Fall UROP3100, Spring

Lack of understanding on corneal response to orthokeratology lens led to a low predictability of the lens efficacy. Ratio of intraocular pressure (IOP) to corneal stiffness was identified as a new factor that influences corneal regression rate after the lens wear. It is hypothesized that cornea with the higher ratio is likely to have less shaping effect and to regress at a faster rate. Ex-vivo porcine eyes (n=??) under 3 designated IOP (5, 15, and 30mmHg) were used to examine the hypothesis. Corneal topography and stiffness were measured before and after the lens wear to monitor the corneal biomechanical changes.

157

School of Engineering Department of Mechanical and Aerospace Engineering

Development of Low-cost Smartphone-based Microscope for Global Health Supervisor: LEE Yi-Kuen / MAE Student: LEUNG Chi San Ivor / PHYS Course: UROP2100, Fall UROP3100, Spring UROP1000, Summer

A new prototype of a low-cost fluorescence microscope is constructed. The prototype comprises a creditcard- size Raspberry Pi computer integrated with a high-quality camera module, commercial microscope objective lenses, long-pass filters, dichroic mirrors, and other components such as 3D-printed parts. Its performance is evaluated and compared with a commercial microscope, Olympus IX 70, in the following aspects: image quality (in both bright-field and fluorescence mode) and point spread function. In addition, the z-axis positioning accuracy and the signal-to-noise ratio of the prototype are measured. Assessment results show that the prototype, with a cost as low as USD $220, performs comparably to the commercial microscope valued at USD $28000, demonstrates high accuracy in z-axis positioning and possesses a high signal-to-noise ratio.

158

School of Engineering Department of Mechanical and Aerospace Engineering

Ambular: An Emergency Response Drone Supervisor: LI Larry / MAE Student: AYDOGMUS Emre Can / AERO KAN Ho Hung / AE SWAMINATHAN Srivatsav / AE Course: UROP1100, Fall UROP1100, Fall UROP2100, Fall UROP3100, Spring

Ambular, an interdisciplinary aviation project conducted in collaboration with the International Civil Aviation Organisation (ICAO), aims to develop new technology for electric vertical take-off and landing (eVTOL) vehicles. Over the next decade, such eVTOL vehicles are expected to play a leading role in the transformation of the transport sector, providing a new mode of direct transport for medical evacuations and paramedics. HKUST is spearheading the aerodynamic design and optimization of the Ambular drone. Our initial testing of the original Ambular design revealed several aerodynamic issues, the biggest of which being the negative lift generated by the fuselage and motor arms. Consequently, the current stage of the project focuses on redesigning the fuselage and optimizing the motor arm geometry. We performed extensive wind tunnel experiments using different fuselage and airfoil designs. The data were systematically collected and analyzed, and were then used to guide the development of a new generation of Ambular components. This report describes the wind tunnel testing and design optimization processes, with a focus on improving the aerodynamic efficiency of our proposed fuselage design for the Ambular drone.

159

School of Engineering Department of Mechanical and Aerospace Engineering

Ambular: An Emergency Response Drone Supervisor: LI Larry / MAE Student: CHOI Tsz Kik / MAE LAU Siu Man / AE Course: UROP1100, Fall UROP1100, Fall

For the Ambular fuselage, we experimentally investigated the performance of two main designs: a double-bubble design, and a hawk design. This report focuses primarily on the first design. Originally, wind tunnel testing and computational fluid dynamics (CFD) simulations had been planned near the end of the semester. However, due to technical and scheduling issues partly arising due to the pandemic, wind tunnel testing could not be performed as scheduled. Therefore, this report focuses mainly on CFD simulations done using Solidworks. The numerical data generated by the CFD simulations will help guide the design of future wind tunnel experiments conducted on the Ambular drone.

Early Detection of Thermoacoustic Instability in Solid Rocket Motors Supervisor: LI Larry / MAE Student: GANI Bryan Suryaraso / MAE Course: UROP2100, Fall

Thermoacoustic instability is an important problem in combustion systems. It may lead to excessive fatigue loading and structural failure of the entire system. Effective and timely control of the phenomenon is necessary to prevent catastrophic failure. However, thermoacoustic systems are highly complex and nonlinear, so it is essential for model-based control methods to have access to a representative system model. In this project, we used the framework of SINDy (Sparse Identification of Nonlinear Dynamics) to discover a model representative of a real thermoacoustic system, namely a Rijke tube driven by a laminar premixed flame. The output of the Rijke tube is simulated and fitted by the SINDy model. In this way, we were able to derive a surrogate model of the Rijke tube, which could then be used in the future to develop new passive and active control strategies.

160

School of Engineering Department of Mechanical and Aerospace Engineering

Study of Flexible Wing Kinematics of a Free-flying and/or Controlled-flying Dragonfly Supervisor: QIU Huihe / MAE Student: YIU Kwan Ho / AE Course: UROP1100, Fall UROP2100, Spring

Given the dragonfly’s wing mechanism has the advantages of the better energy efficiency, higher maneuverability and higher agility, the concept of the flapping wing has the potential to be put into the future flying machines. This study analyzes the kinematics of a controlled-flying dragonfly under side-wind conditions by observing the pattern of the generated aerodynamic force. From the experiment, it is found that the section lift coefficient of the wings on the side-wind side is lower than another side. Furthermore, the section lift coefficient of the forewings decreases with the stronger side wind, but the section lift coefficient of the hindwings does not vary with the side wind strength.

Environmental Impact by Air Traffic: Assessing Aircraft Noise nearby HK Airport Supervisor: REDONNET Stephane / MAE Co-supervisor: HORNER Andrew / CSE Student: CHAN Ngai Nam / AE Course: UROP2100, Fall UROP3100, Spring

This UROP work consisted in performing field test measurements of actual aircraft noise to validate a pre-existing aircraft noise prediction algorithm developed at HKUST. To this end, the noise levels from various departure and approach flights were measured in two locations of Hong Kong city. This was achieved through three consecutive phases, namely 1) preparing the experimental acquisition, 2) conducting the actual field testing and 3) post-processing the noise signals. A total of 60 flights were recorded, with 6 different aircraft types considered. The noise data were then post-processed through specific manipulations, to derive specific relevant metrics, e.g. Sound Exposure Level (SEL), 1/3-octave frequency spectrums of SEL, Sound Pressure Level (SPL), etc. The results were then compared to those coming from the aircraft noise prediction algorithm, with a special focus put on 15 specific flights (mixing both A330-300 and B777-300ER aircraft). Comparison proved to be favourable enough, with both predicted and recorded noise levels falling in an acceptable range of a few dB (5 dB at most). This is encouraging if one considers how challenging it is to perform aircraft noise field testing with limited experimental means as the ones used here. This work was presented at the AIAA Region VII Student Conference 2021.

161

School of Engineering Department of Mechanical and Aerospace Engineering

Exploring the Aero-Science behind Flyingsuits Design Supervisor: REDONNET Stephane / MAE Student: SACHDEVA Nidhish / SENG Course: UROP1000, Summer

This UROP 1000 project constituted a first step towards the computational exploration of the design of wingsuits (also called flying-suits). To this end, the computational strategy was first established, being then validated through a popular benchmark case. This consisted in performing various steady aerodynamic simulations involving a NACA0012 airfoil, which was achieved by solving the steady RANS (Reynolds Averaged Navier-Stokes) equations along with a SST k-ω turbulence model using a structured mesh. The lift and drag polars obtained compared favourably with results coming from the literature. Upon its validation, the computational methodology was then applied to the case of a 3D wingsuit model, delivering early insights about the complex aerodynamic flow features it entails. A follow-up UROP project is on its way.

Digitalization of Wet Lab Project Supervisor: YANG Jinglei / MAE Student: LAU Yuen Hei / MAE Course: UROP1100, Fall

During this semester, I worked on a UROP project supervised by Prof Jinglei Yang in the MAE department. The goal of this project is to digitalize chemical storage inside wet labs with an in-house innovated X-Tray system that detects and measures the weight of chemicals with time stamp and user ID. The information will then be stored in a database to track the chemical usage in the lab. The tray could also warn users when dangerous chemicals were put together to minimize chemical accidents. In this project, I improved the design of X-Tray in both mechanical and software aspects. For mechanical design, I optimized the accuracy of weight sensing & reading and rigidity of the system by changing the design and configuration of weight sensors. For software, I designed a more user-friendly GUI which can visualize the data obtained from the system with the help of another research assistant. This report will be focusing on the design of the hardware and analysis of the improvements. Through this UROP project, I have gained a lot of experiences to solve practical problems via learning-by-doing engineering mindset and our “Can Do” spirit. I do appreciate very much this opportunity provided by HKUST and Prof Yang.

162

School of Engineering Department of Mechanical and Aerospace Engineering

Smart Cushion Supervisor: YU Hongyu / MAE Student: WUN Ga Yun Esther / MAE Course: UROP1100, Fall UROP2100, Spring

The smart cushion system is a system integrating the sensors into a cushion to monitor the sitting habits of people. Two different designs of the substrate were suggested, one adopting a folding-stretching method for a flexible printed circuit board substrate, one using a Miura-Ori substrate design with sensors underneath. Both designs aim to minimize the number of sensors, the material usages and the production costs, while maximizing the coverage area. This paper would focus on the stretching machine for the folding-stretching method, as a continuation of the previous report, as well as the simulation results of the Miura-Ori substrate design.

163

Abstracts of UROP Projects 2021-22 School of Business and Management

School of Business and Management Department of Accounting

Department of Accounting Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: CHENG Bo / IS Course: UROP1000, Summer

This study aims to investigate the accuracy of the FinBERT model, a domain-specific pre-trained deep learning NLP model, in classifying the Environment, Social, and Governance (ESG) categories from companies’ ESG reports or Sustainability reports. After manually labeling sentences from the annual reports into respective ESG categories based on the MSCI ESG rating framework, 4,500 sentences from 55 companies in 11 industries are used as samples to proceed with model fine-tuning and testing. Two Bidirectional Encoder Representations from Transformer (BERT) models are chosen to train and compare their performances: the base BERT model (cased) and the FinBERT model. The overall findings suggest the fine-tuned FinBERT model outperforms others for the ESG classification accuracy though the alternative pre-trained BERT model produces qualitatively similar results.

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: GONG Xiuyi / FINA Course: UROP1100, Summer

Bert has been widely recognized as a state-of-the-art pre-trained model of natural language processing. Finbert as an advanced version of NLP in the Financial area has also achieved ideal performance in tasks related to financial texts.

In this report, an attempt was made to further finetune Fin-bert on an ESG-themed dataset to check its performance. The dataset is made up of sentences randomly selected from the ESG report of S&P 500 companies and labeled with one ESG theme. The NLP task is text multi-class classification with 9 classes of different ESG focus. And comparisons were also made among five other models including Bert, Random Forest, Support Vector Machine, Naive Bayes, and Long short-term memory.

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: HE Zhiyuan / MATH-SF Course: UROP1000, Summer

In this paper, I used a dataset with sentences from ESG reports testing the performance of the pretrained finBERT model in classifying the sentences related to finance. The following is the main structure of the paper. At the start, I test the performance of finBERT model under different values of four hyperparameters, and discuss the possible reason behind the behaviour of the model. Then picking the best-performing values of all four hyperparameter to produce the best finBERT model in this classification task. In the last part, I test the performance of finBERT model while training dataset is severely unbalanced.

166

School of Business and Management Department of Accounting

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: KHURANA Vinayak / CPEG Course: UROP1000, Summer

In this paper, a performance evaluation of five machine learning algorithms was conducted with regards to corporate ESG text classification. Using a sample of 4500 manually labelled sentences from ESG and sustainability reports across 11 industries, it is shown that the Support Vector Machine (SVM) was the optimal algorithm of the other four models including Naïve Bayes, XGBoost, Google’s bidirectional encoder representation from transformers (BERT), and HKUST’s FinBERT, an algorithm customized from BERT for financial text. Two performance metrics were chosen to evaluate performance involving the accuracy score and execution times. Although BERT achieved the highest accuracy and Naïve Bayes achieved the fastest execution times, SVM was chosen due to its optimal performances with regards to each metric.

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: LI Qiuru / IS Course: UROP1000, Summer

In this paper, we use a finance-domain pre-trained deep learning NLP algorithm – FinBERT to fine-tune for an NLP classification task and compare the performance with another NLP algorithm – BERT. First, we collected ESG reports from various companies in different industries to extract financial sentences and labelled the sentences based on 13 categories in ESG, including non-ESG and non-sentence with our own knowledge. Then, we randomly selected 4500 sentences for FinBERT fine-tuning. In the fine-tuning step, we used the pre-trained FinBERT and BERT model to fine-tune for the ESG-sentences classification task and did the comparison of the accuracy, we found that FinBERT performed slightly better than BERT. Our results have implications for those who want to extract insights from financial texts.

167

School of Business and Management Department of Accounting

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: MORSI Mohamed Sobhy Mohamed Hassan / SENG Course: UROP1000, Summer

This study aims at fine-tuning the domain-specific pre-trained model FinBeRT to create a multi-class text classifier in the finance domain. Further, the results were compared against various other NLP models, namely the uncased base BeRT model, RoBeRTa model, Support Vector Machine Classifiers, Decision Trees, Random Forests, Logistic Regression, and Naive Bayes Classifiers.

Experiments have shown that the BeRT family of models outperformed the other methods, with minor nuances in test accuracy and robustness.

The main task given to these models was to classify ESG sentences into 9 predefined categories according to their meanings.

The code and datasets used in this report are available on the following GitHub fork: https://github.com/mohamedsobhi777/FinBERT

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: OUYANG Xiangting / SBM Course: UROP1000, Summer

In recent years, ESG ratings have been arousing the attention of investors and driving their need to acquire related information. As manually analyzing some financial information may be labor consuming, there are some studies on financial sentiment analysis based on deep learning NLP algorithms. To effectively perform the analysis of ESG related sentences, we apply FinBERT to compare its performance of classification task with BERT. We prepare a sample of sentences from ESG reports of different firms and label them according to the MSCI ESG Ratings Methodology. Contrary to our prediction, BERT performs slightly better than FinBERT in ESG classification, which may not suggest superior performance of deep learning NLP algorithms that include domain-specific knowledge (FinBERT) than those rely on general texts (BERT) in a specific domain.

168

School of Business and Management Department of Accounting

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: SHE Fong Wing / RMBI Course: UROP1100, Summer

In this paper, the performance of FinBERT, the Large Language Model that adapts to the financial domain, on multi-class classification of ESG categories based on the MSCI framework is evaluated. ESG sentences that are used in this research are extracted from the CSR reports of companies in 11 different industries. The evaluation is based on comparisons with other 11 Deep Learning and machine learning algorithms, including Naïve Bayes, Logistic Regression, Linear SVM, Random Forest, MLP, CNN, LSTM, Bi-directional LSTM, GRU, and BERT. Through fine-tuning of all the mentioned models, FinBERT is found giving the best performance in the task, but still have its performance extremely close to BERT, and the per-class precisions of FinBERT are less stable than BERT.

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: XIAO Yicong / QFIN Course: UROP1100, Summer

This project fine-tuned a pre-trained FinBERT model for classifying financial text in ESG context into 9 predefined categories. The accuracy was then compared with those obtained by training and testing 4 Bag-of- Words based algorithms including Logistic Regression, Random Forest, Linear Support Vector Classification, and Naive Bayes, as well as 2 deep learning models including BERT and LSTM on the same data set. The project found that the deep learning models selected, which process text sequences instead of word counts, consistently outperformed Bag-of-Words based algorithms. BERT and FinBERT were also found to outperform LSTM and showed consistency in accuracy when reducing the size of the training set. No significant difference in performance was found between BERT and FinBERT.

Deep Learning in Natural Language Processing Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YANG Yi / ISOM Student: ZOU Xinying / ACCT Course: UROP1100, Summer

In the UROP project, we conducted data collecting by gathering ESG reports of companies in various industry with different scales from the internet, followed by data matching work that assigns sentences in those reports into different ESG labels. To make a preliminary analysis on the matching data, we compared the labeling result done by two different students and adjusted the data to get a more reasonable label through discussion in the weekly meetings. Finally, we did the fine tuning on FinBERT (BERT for Financial Text Mining) to classify ESG discussions.

169

School of Business and Management Department of Accounting

Investment Analysis with Machine Learning Supervisor: YOU Haifeng / ACCT Student: LIU Zhetian / MAEC Course: UROP2100, Fall

This report investigates the possibility of predicting Chinese industries’ returns using cross-industry historical information in technical analysis. Both linear and non-linear machine learning tools are used to generate predictions, and it is highly possible that such predictability exists based on out-of-sample forecasting results, with an average improvement of 47%. We do not find evidence that OLS Post − LASSO can improve OLS’s predicting performance, which contradicts existing literature. Possible applications, improvements, and generalizations of our study are also analyzed.

Investment Analysis with Machine Learning Supervisor: YOU Haifeng / ACCT Student: YEUNG Man Yin Michael / MAEC Course: UROP1100, Fall UROP2100, Spring

The advent of machine learning facilitated the extraction of sentiment from financial text corpus. Recent literatures have introduced state-of-the-art models to make stock return predictions. In this study, replication of FarmPredict model by Fan et al. (2021) is implemented to examine its ability to mine sentiment from Chinese analyst reports and make Chinese stock returns predictions. The framework implemented with grid-search and cross-validation can reach optimal test Spearman correlation of 11.7% comparable to that found in the SESTM framework (11.8%) in previous work (Yeung, 2021). Variant of the framework with input not transformed to word embeddings have better performance. The magnitude of correlation provides evidence that FarmPredict can generate predictions useful for investment analysis but to a similarly limited degree.

170

School of Business and Management Department of Economics

Department of Economics Projects on International Trade, Innovation, and Corporate Finance Supervisor: GONG Robin Kaiji / ECON Student: LAM Lai Man / ECOF LEUNG Ho Fung / ECOF WONG Kelly / ECON ZOU Xinying / ACCT Course: UROP1100, Fall UROP1100, Fall UROP1100, Fall UROP1100, Fall

In the UROP project, we conducted data cleaning and preliminary analysis of the IPO matching and entity list matching data. In extension to the project, this paper investigates the impact of the US-China trade war on both countries’ economies. It is found that the trade volumes of both countries decreased. The underlying factors boil down to the relocation of productions to other Southeast Asian countries and the reduction in investment in technology. It is also found that China's retaliatory tariff caused a decrease in US’s export value. Additionally, the US tariff resulted in a complete pass-through into US domestic prices. This paper concludes that both countries experienced a worse-off economy thanks to the trade war.

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: CHAN Cheuk Yin Rudolph / ECON YU Yue / MAEC Course: UROP2100, Fall UROP3100, Spring UROP1100, Fall UROP2100, Spring

In our research, we have explored the export premium of Chinese manufacturing firms. By utilizing the firm production survey data and Chinese customs data, we processed firm-level information and sort out the single product firms. We then adopted De Loecker and Warzynski (2012)’s method to estimate firm-level markup and firm-level productivity. After constructing the performance measures, we regress markups on export status and controlled for marginal cost by adding TFP into the regression, we also decompose the export status into entry and exit effects. Our results showed that exporters charge on average higher markups and firms’ markups increase upon export entry.

171

School of Business and Management Department of Economics

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: CHEN Yi / ECOF Course: UROP1100, Fall UROP2100, Spring

In UROP1100, we examine the effects of Chinese retaliatory tariffs on 5G imports under the US-China trade war, which began in 2017. Following the progress, we use a reduced-form difference-in-difference framework to investigate the direct and indirect effects on the HS-6 level. We find that the retaliatory tariffs have a negative though not significant impact on China's imports of five G-related products from the US. Furthermore, its indirect effect on five G-related imports from the rest of the world is also negative and even significant.

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: CHOW Wing Shan / ECOF LAI Cheuk Lam / ECOF Course: UROP1100, Spring UROP1100, Spring

The COVID-19 pandemic has struck the global economy since 2020. As governments started to impose social distancing measures and lockdowns, it affected global supply chains and hindered economic growth with rising unemployment. This paper investigates the effect of the pandemic on the import patterns between Hong Kong and its major trading partners. We have first identified a positive correlation between confirmed COVID-19 cases and import values at the aggregate level based on the gravity model. We also applied the gravity model to product-level data, which showed a mixed result across industries. The rise in aggregate level import values over time was mainly driven by the rising demand for electronic gadgets induced by trends in remote working and the availability of digital services. However, goods such as optical equipment and meat showed a significant negative correlation with COVID-19 cases in our results. We expected an uneven recovery across sectors in the post-covid era.

172

School of Business and Management Department of Economics

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: GUO Liangyu / ECOF YANG Jingdian / MAEC Course: UROP1100, Fall UROP2100, Fall

Since its launch in 2013, China’s Belt and Road Initiative (BRI) has enabled the Chinese government to actively invest in nearly 170 countries globally with a focus on large-scale infrastructure construction projects and energy goods. This progress report presents how the Belt and Road Initiative affects China’s different types of trade and FDI patterns at aggregate and sector levels. We found that at the aggregate level, countries receiving a higher amount of FDI from China also have a higher level of trade flow with China, after controlling the time fixed effect and country characteristics. For the energy FDI and manufacturing FDI, both of them have a significant impact on aggregate trade after controlling for time fixed effect and country characteristics. However, as shown in different regressions we conducted, there is mixed evidence for the incremental impact of being an initially targeted BRI country on trade value.

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: HE Jiawei / ECON WANG Yingfan / MAEC Course: UROP1000, Summer UROP1000, Summer

This article studies the changes and impacts of the China-United States trade war in the context of the global COVID-19 pandemic. We first selected the data of imports from China and the U.S. in the last twenty months and did descriptive statistics and analysis. Next, we predict that there is an inevitable link between the increase in tariffs and the import and export volume between the two countries and conduct calculations and analysis on a large number of data to test our hypothesis.

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: LEE Daye / MAEC OUYANG Yuxuan / MAEC Course: UROP1000, Summer UROP1000, Summer

Innovation is widely regarded as one of the most substantial factors for companies to continuously operate and develop. Our research investigates whether credit constraints enhance or deteriorate the firms’ innovation. We use the firm-level production data, the annual surveys of Chinese manufacturing firms conducted by the National Bureau of Statistics of China (NBSC) in 2004 to capture firms’ innovation while we obtain factors to proxy credit constraints from Monova (2015) and Fan (2015). In this progress report, our main identification is that research showing the relationship between credit constraints and innovation is worthwhile.

173

School of Business and Management Department of Economics

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: LI Yuhe / RMBI Course: UROP2100, Fall

To investigate the correlation between Foreign Direct Investment (FDI) and domestic firms’ productivity, this paper has the following three focuses, data cleaning, horizontal and vertical backward regression. We first dropped inappropriate observations from the CIE data, then converted the annual cross-sectional data into panel data. Then, we followed the standard approach by Jarvocik to calculate the horizontal and vertical backward proxies, with the use of the 2002 Chinese input-output table. For horizontal spillover effect, We obtained a mixed result of both negative and positive horizontal effects when adding different fixed effects compared to the negative horizontal effect in the previous works. For vertical backwards spillover effect, positive and significant coefficient was found, indicating the positive productivity spillover effect under foreign presence in the downstream industry.

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: LIAO Caixing / ECOF SU Xuanjing / ECOF Course: UROP2100, Fall UROP2100, Fall

Following up on the aggregate level analysis in UROP 1100, this report proceeds to investigate whether there exists a trade diversion in China caused by the trade war starting from 2018. Based on monthly data of China’s export and import at the HS6 product level, this report adopts a reduced-form difference-indifferences framework in analysis. We separately study Chinese export diversions using US sanction lists and Chinese import diversions using China retaliation lists. We find significant trade diversion effects using some specific tariff lists but not all. Such heterogeneity in lists will be examined in future work.

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: LUK Wing Yiu / ECOF Course: UROP2100, Fall UROP3100, Summer

This study focuses on the intra- and inter-industry productivity spillover by inward foreign direct investment (FDI) in China. Using the firm-level CIE panel data, I found the existence of positive productivity spillover by foreign investments excluding those from HK, Macau and Taiwan through backward and forward linkages. Lagged forward proxy is proved to be more predictive and accurate as compared to the contem poraneous proxy. Foreign participation in the same industry poses negative impact on the productivity growth of their domestic counterparts. Lastly, by separating the backward proxy, fully-owned and export-market oriented foreign firms are found to have positive impact on domestic productivity growth.

174

School of Business and Management Department of Economics

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: WANG Kaiwen / MAEC Course: UROP1100, Fall

This report documents the trade barriers induced by the United States and China in the context of trade conflicts between the two countries since 2017. With special interest on the effect of trade policies on 5G products, we aim to employ regressions of Chinese import level on Chinese retaliation tariff as well as Chinese export level on US tariff based on the data covering the period from January 2016 to December 2019. This HS-6 level analysis so far (1) has not sufficed to provide evidence for significant change in Chinese import level in response to retaliation tariff, and (2) confirm that all five-G related products and homogeneous unrelated products are exposed to US tariff. This latter finding also leads to documentations on facts regarding non-tariff measures imposed by the United States on 5G products.

Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: XU Runzhao / MAEC ZHENG Yufei / MAEC Course: UROP2100, Summer UROP1100, Summer

We aim to study the effect of state-owned enterprises reform in China. The reform started from 1978, with progressive measures. We use the firm-level production data of Chinese manufacturing firms from 2000 to 2007 to observe the changes of enterprises’ performance after the reform and try to analyze the treatment effect of the reform on SOEs compared to non-SOEs. We use the ratio of capital funds from state to define SOEs and find that the number of SOEs decreases, one of the reasons is the privatization of SOEs. Meanwhile, taking per labor output and per firm output as proxy variables, the average output value of each SOEs increases faster than non-SOEs. As a result, the state advanced while the private sector retreated.

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: CHAN Wai Chung / MAEC Course: UROP1100, Summer

This report describes my work in the Undergraduate Research Opportunity Program in 2022 Summer in the project Local Impacts of Belts and Roads Initiative(BRI) Projects. I carried out data collection and cleaning and Geo-coding work in the first stage. Overall, I have finished the data collection and cleaning as well as their Geo-coding for more than 900 datum. For most of projects, I provided a clear commitment, implementation start and completion year and their exact location expressed in ”Latitude, Longitude”. And with the supplement of data from DHS Program, I conducted a simple linear regression analysis of the BRI Projects on the local employment. The results of the analysis shows a negative relationship between exposure to BRI project during window period and employment. While it also shows a very little relationship between exposure and unskilled status.

175

School of Business and Management Department of Economics

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: CHEN Ziji / MAEC Course: UROP1100, Fall

There are thousands of bidding documents in each country and there are also many countries in the BRI projects. However, in a large number of documents, many useless documents are being wrongly selected by the computer which has no relationship with BRI projects. For example, we want some documents containing “南非”, which is a country in Africa but thousands of documents which containing “湖南非物质 文化遗产” or “河南非遗” will be automatically filtered in by computer. Therefore, we need to find out these kinds of wrong documents and delete them. Since the important information in documents is usually not convenient to find, we should also help sort out the information by classification.

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: HUANG Shi / ECON Course: UROP1100, Summer

Since China launched the Belt and Road Initiative (BRI) in 2013, the related regions and countries have been highly impacted by the investment in construction and worker mobility. Several empirical research has focused on the role of BRI projects in economic development, but there are still many aspects to explore. This report mainly focuses on the data processing work of the research, which focuses on the BRI project, including data matching, geolocation matching, and project locating. It is still on the way to having a clear result of the impact. Through the data of the location and matched projects, more studies on the impact of BRI can be conducted.

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: JIANG Yunshu / ECOF Course: UROP1100, Fall UROP2100, Spring

This project is comprised of two parts. The first part investigates the way that Chinese enterprises engage in the Belt and Road Initiative. The Belt and Road Initiative offers Chinese companies an opportunity to engage in the international market. Through participating in international competition, the competitiveness of enterprises could be enhanced, winning a broader space for long-term development. By analyzing the filtered bidding data, we can find out the distribution of enterprises, for example, their industries. At the same time, we could get to know the amount of money they invest. The second part of the research is to study the local impacts of the BRI projects. By collecting and geocoding China’s foreign aid project, we can know the detailed distribution of BRI projects. By digging deep into the statistics of local economic development and other relevant data, for instance, employment statistics, we could get to know the change of the welfare of local people, therefore, making feasible suggestions on how to enhance the life quality of local people.

176

School of Business and Management Department of Economics

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: JIANG Yuxuan / SSCI Course: UROP1100, Summer

BRI (The Belt and Road Initiative) is a granted global infrastructure plan initiated by the Chinese government in 2013. Chinese Prime Minister Xi Jinping issued it as a “Major Country Diplomacy” strategy. Up to August 2022, 149 countries have joined BRI. Our team is working on The Local Impacts of BRI Projects from different perspectives (e.g., trade, industrial activities, firms’ activities) and finding out both the intended and unintended consequences. In this summer (6th July–31st August) UROP project, we focused on matching the geo location manually and used R code to scrape the geocode to make an intuitive map. Professor Yatang Lin and Jin Wang supervise our team. And Ph.D. candidates Pak Hung Lam, Fangyuan Peng, and M.Phil. student Mingjie (Abel) Xu co-supervise our project.

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: LI Tianyu / ECOF Course: UROP1100, Spring UROP2100, Summer

This report mainly includes the processes and results of my tasks. Also, the report will illustrate each step’s coding and respective reasons for it. The project is about collecting information related to the projects in Bridge and Road Initiative (BRI) and using Python to do geo-coding to analyze the local impact of them. Each student is assigned to several countries’ BRI projects and through searching, scrapping and organizing data, they can manage to do final matching task by using the aid-data to find out the matching results. After that, it will illustrate the manual geolocation search of some unmatched results. Then, R are applied for the mapping for these projects to see the location and frequency of my assigned data. Finally, the report will briefly explain some significant projects for each countries.

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: LI Xinzi / ECOF Course: UROP1100, Fall

This study will analyze the implications of BRI from firm-level aspect to investigate how Chinese enterprises are linked by BRI. The study will also explore the incentives for Chinese governments and firms to participate in overseas projects. We assume Chinese companies, especially manufacture factories, are connected under the BRI projects via bid invitation and bidding. We will compile a geocoded firm-level project bidding dataset to examine and find out more information. This research is still in progress, the task of this semester is to clean the bidding data of corresponding BRI countries.

177

School of Business and Management Department of Economics

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: LI Ziyuan / ECOF Course: UROP1100, Summer

This report documents the work involved in merging raster data on 7 development-economic, environmental, and geographical indicators onto a spatial dataset of 5km-by-5km grids in Africa. The number of years we merge for each variable ranges from 1998 to 2020 (the span differs slightly between variables). The report also details the retrieval of trade flow records between 24 African and 230 world countries-a total of 5520 pairings-from UN's Comtrade Database. We expect to download a total of over 20 million observations and are currently over half-way through. Our preparation work here allows us to have comprehensive data ready for analysis and a global grid associated with local variables of interest on which to designate BRI projects as geographical study units.

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: WAN Yuet Lung / ECOF Course: UROP1100, Spring

The Belt and Road Initiative was launched in 2013 with the aim of enhancing regional integration, increasing trade and stimulating economic growth. In these 9 years, plenty of infrastructures were built and fostered the development of the local economy. The roads and railways constructed were significant symbols of the initiative, which represent the connectivity of the “silk road”. In this project, we have discovered the roads and railways constructed by the initiative in Africa. Also, we summarized the data in searching for the distribution of projects within the region. 139 routes are covered in this project, attributed to 31 different countries.

The Local Impacts of BRI Projects Supervisor: LIN Yatang / ECON Student: ZHANG Han / ECOF Course: UROP2100, Fall

Since its launch in 2013, the Belt and Road Initiative (BRI) has brought a tremendous impact on major regions. An increasing amount of research is centring on the environmental impact of BRI, given its unprecedented implementation in infrastructures. This article studies the two perspectives on the effect of BRI regarding green development and sustainability. Some research suggests BRI could accelerate the green development in its participating countries, while others raise concerns over its adverse environmental impact such as increased carbon emission. This article also suggests potential research areas once more detailed information on BRI projects are collected.

178

School of Business and Management Department of Economics

Kin 'Taxes'? How Filipina Domestic Workers in Hong Kong Manage Family Expectations Supervisor: VISARIA Sujata / ECON Student: TANG Tsz Yeung Alvis / ECOF Course: UROP2100, Fall UROP3100, Spring

Filipina domestic workers in Hong Kong earn higher incomes than in the Philippines but most report no improvement in their living standards. Continuing on previous fieldwork, the research team conducted a mixed-methods study with 25 participants this past semester. In this report, I highlight three emergent themes from my reading of 5 qualitative in-depth interviews. Firstly, it is common for domestic workers to view migration as an intra-household venture. Secondly, some domestic workers avoid loans. Last but not least, the Christian faith has had a rather large influence on some participant's financial life and beyond. I also discuss potential future directions and interventions.

Exploratory Research to Inform a Financial Planning Tool for Migrant Domestic Workers in Hong Kong Supervisor: VISARIA Sujata / ECON Student: NAM Yebin / ECON PURI Sneha / ECON SONG Xiaolong / ECOF Course: UROP1100, Spring UROP2100, Fall UROP3100, Spring UROP1100, Spring

The project, “Exploratory Research to Inform a Financial Planning Tool for Migrant Domestic Workers in Hong Kong'', aims to conduct exploratory study on the financial wellbeing of Filipna domestic workers in Hong Kong (commonly referred to as OFWs) and feed into the design of a new financial planning tool to improve their migration outcomes. While this Undergraduate Research Opportunities Project (UROP) is a continuation of the studies done in previous semesters, this particular semester we conducted interviews of 25 Filipina domestic workers (OFWs) in Hong Kong and are at the stage of quality control and data cleaning.

Exploratory Research to Inform a Financial Planning Tool for Migrant Domestic Workers in Hong Kong Supervisor: VISARIA Sujata / ECON Student: ONGRATANAPHOL Ramon / MAEC Course: UROP2100, Fall

Filipina domestic workers (commonly referred to as OFWs) make up an integral part of Hong Kong’s care economy. There are around 170,000 OFWs in Hong Kong. We conducted quantitative surveys with 57 OFWs in July and August 2021. The first stage of this process was to post an advertisement on the Facebook page of a Filipino newspaper that is popular amongst OFWs, The Sun. To express their interest and willingness to be interviewed, we asked them to complete a 5-minute-long survey about their life (such as marital status, number of kids, age, etc), based on which we could divide them in Focus Groups, and contact information. After they were divided into appropriate groups, we contacted them on a first-come-first basis.

179

School of Business and Management Department of Finance

Department of Finance Trade Shows and Supply Chain Formation Supervisor: BISETTI Emilio / FINA Student: CHEN Jinghui / FINA XU Suixin / QFIN YUAN, Wai Man / FINA ZHANG, Jiayi / MAEC Course: UROP1100, Spring UROP1100, Fall UROP2100, Spring UROP1100, Spring UROP1100, Spring

Our research focuses on the field of corporate finance in the United States. We mainly use a large amount of retail data to determine the competitive relationship among companies by matching their products and giving a score based on the product matrix. In order to access the appropriate data, we have done some basic data cleaning jobs, using Python to match firm names from different databases and do manually matching. We also use inner functions in Excel to sort the company information into categories. The above methods can be utilized for broader company scope and longer time horizon.

Measuring Product Market Competition Supervisor: BISETTI Emilio / FINA Student: CHEN, Jinghui / FINA HE Jinwen / COGBM HU Yumeng / SBM HUANG Jiawei / IEEM LEUNG Suet Ying / FINA LO Kim Wa / RMBI PHAM Thanh Hai Binh / IPO XU Suixin / QFIN YUAN Wai Man / FINA Course: UROP2100, Summer UROP1000, Summer UROP1000, Summer UROP1000, Summer UROP1100, Summer UROP1000, Summer UROP1000, Summer UROP3100, Summer UROP2100, Summer

The project aims to construct a firm-level competition measure based on product information from detailed scanner data. We obtain product bar codes and sales revenue from Nielsen and link back to the public firms in the Compustat database. Following the literature on Text-based Network Industry Classifications (TNIC), our research makes use of category division and market shares to build up the product category matrix of each company and prepare the data to calculate similarity scores between the two companies.

180

School of Business and Management Department of Finance

Risk-management Duel: Financial vs Operational Risk Management Supervisor: MACKAY Peter / FINA Student: CHIU Kam Wan / MAEC LIU Chengcheng / SSCI LUO Xiaonan / DSCT PANG Kai Fun / IS XU Tunan / RMBI Course: UROP1000, Summer UROP1100, Spring UROP1000, Summer UROP1100, Spring UROP1000, Summer UROP1100, Summer

The operations of companies ordinarily balance the operational risk management and financial risk management through preserving or spending their asset. Generally, companies alter their company scaling based on their incoming, total asset or total liability. This project is aiming to analyse financial and operational management dataset of several mining companies and reveal some inconspicuous relation between operational risk management and financial risk management. The dataset includes data over the past 20 years representing some property of mines such as proven reserves, probable reserves and indicated or measured reserves for each mine operation. Data will be analysed using some hypotheses testing methods to determine how a company’s portfolio (of mines) implies how it deal with financial risk. At this stage, data of all companies have probably already been completely collected, and will be prepared for more testing and analysing to reach to the conclusion of this research project.

Family Office Research - Understanding the Trend of Wealth Supervisor: PENG Qian Winnie / FINA Student: LI Tin Yau / FINA MA Ka Hei / FINA Course: UROP1100, Spring UROP1100, Spring

In recent years, the number of billionaires has grown rapidly worldwide. Few academic papers and business reports have been published to examine the characteristics and demands of billionaires. In this research paper, we focus on studying the Chinese billionaires, as the numbers have surged in recent years. Based on an in-depth analysis of the billionaire lists and the review of market data, we seek to offer a more comprehensive understanding of the source of wealth of Chinese billionaires and the possible future trend. We find that the ranking of the billionaires has changed significantly in recent years. Some new economy are rising up while the importance of the real estate industry, which used to be the dominant source of wealth for Chinese billionaires, has declined.

181

School of Business and Management Department of Finance

The Economics of Data in Financial Markets Supervisor: XIONG Yan / FINA Student: HONG Zhaoyang / QFIN Course: UROP1100, Fall

In the process of participating in this project, I mainly completed two aspects of work. They are the research of the voluntary disclosure of corporate site visits and the hand collection of IPO data. About the research of the voluntary disclosure of corporate site visits, we focus on the observability of investors' information acquisition. Usually, people's prior is that funds are very secretive about their research activities. So, we want to study examples of voluntary disclosure of company field visits by funds. About the hand collection of IPO data, our research topic is a new regulation-testing the waters and we want to explore its implications for the market. We studied this act, especially the influence of water testing on the market.

The Economics of Data in Financial Markets Supervisor: XIONG Yan / FINA Student: ZHANG Zirui / ECOF Course: UROP1100, Fall

In the U.S., a decrease in the number of the public companies has attracted high-level concern from regulators, prompting changes to the IPO process to welcome more young, high growth firms (the JOBS Act of 2012). However, there is always a controversy about the impact of the JOBS Act on the IPO processes and results of small but high growth enterprises. To have a better understanding of the specific impact the JOBS Act had since the decree went into effect, we selected 208 small companies and focused on their S-1 financial reports and made statistical analysis on relevant data.

Institutional Investors and Environmental, Social, and Governance (ESG) Preferences Supervisor: YEGEN Eyub / FINA Student: CHAWLA Anhad Singh / SUSEE Course: UROP1100, Fall

This report analyses ESG-related shareholder proposals from institutional and individual investors to see their ESG preferences. Shareholder proposals are a powerful tool to raise and communicate issues to management, with the potential to drive change. Proxy proposals and data is collected from SEC’s EDGAR and ISS databases to conduct analysis. Evaluation of the data reveals majority of proposals come from institutional funds that act or are aligned with public interest (SRI, religious funds, pension funds, etc.). The Board almost always recommends voting “Against” ESG-related shareholder which creates doubt on the effectiveness of such proposals, but they are nevertheless excellent ways to raise management awareness via presentations in Annual General Meetings. Most reports ask for additional disclosures and transparency, while the minority ask for major changes in company operations. Future trends regarding shareholder proposals relating to quantity, type, and blockchain applications are discussed.

182

School of Business and Management Department of Finance

Institutional Investors and Environmental, Social, and Governance (ESG) Preferences Supervisor: YEGEN Eyub / FINA Student: CHOW Wing Shan / ECOF Course: UROP1100, Fall

Traditionally, environment, society and governance (ESG) issues were not the first priority to the investors. However, in recent years, more and more investment strategies incorporate ESG considerations. The ultimate goal of the project is to examine whether certain types of institutional investors, ranging from mutual funds to endowments, have specific preferences to ESG practice. We studied 5,500 ESG-related shareholder proposals in the proxy statements, which were provided by the U.S. listed companies regulated by U.S. Securities and Exchange Commission. We identified the main ESG areas that institutional investors proposed, and the responses from the Broad of directors. We were interested to understand the correlation between institutional investors and ESG preferences, as well as how the management level of the public corporations incorporates the ESG considerations suggested by the investors.

Institutional Investors and Environmental, Social, and Governance (ESG) Preferences Supervisor: YEGEN Eyub / FINA Student: GU Zhaoyi / QFIN Q IAO Yanliu / QFIN Course: UROP1100, Spring UROP1100, Spring

In this report, we are primarily concerned with the ESG investment preferences of two financial institutions: BlackRock and Fidelity. Our research methodology consists mostly of case studies, qualitative readings of corporate reports, and quantitative analysis of publicly available data. We discovered that both BlackRock and Fidelity are interested in the Asia Pacific market. Furthermore, BlackRock has done the best in the stewardship engagement, with its emphasis on strategy's materiality and transparency and quantitative ESG metrics including decarbonization, diversity, independence, and etc. Fidelity, on the other hand, has been actively involved with firms that are working to eliminate modern slavery and promote clean energy. decarbonisation

Institutional Investors and Environmental, Social, and Governance (ESG) Preferences Supervisor: YEGEN Eyub / FINA Student: LAM Wai Wang / FINA Course: UROP1100, Fall

Nowadays, Environmental, Social, and Governance (ESG) is no longer just an ethical issue, but it also becomes one of the key factors being considered by the institutional investors. In fact, ESG issues, particularly the environmental concerns, have been gaining more and more attention around the globe because of the extreme weather events induced by the climate change. On the other hand, the sustainable development goals set by the United Nations also arouse investors’ interests in the sustainability behaviours of the companies they invest in. However, these judgement could be just noise without statistical support. In this research, Close to 1,200 Shareholders proposals proposed by the institutional investors that are related to ESG issues would be studied to understand their preferences in different ESG fields of the corporates.

183

School of Business and Management Department of Finance

Institutional Investors and Environmental, Social, and Governance (ESG) Preferences Supervisor: YEGEN Eyub / FINA Student: LI Jingrui / MAEC MAK Hoi Yung / MARK NG Josiah Jian Wei / SENG ZENG Ziyu / IPO Course: UROP1000, Summer UROP1000, Summer UROP1000, Summer UROP1000, Summer

This paper mainly explores the institutional investors and their environmental, social, and governance preferences by researching different literature in the ESG field and studying in-depth the leading enterprise in ESG and financial field, our team got a basic idea of institutional preference on the criteria of ESG. The team also found that companies with higher ESG commitment and conduction will have better long-term performance in both financial and non-financial fields, and also get a higher long-term value. There are still some challenges we meet during our research, some further research is still needed as this is quite a broad topic that can include all works of life.

Institutional Investors and Environmental, Social, and Governance (ESG) Preferences Supervisor: YEGEN Eyub / FINA Student: SUN Wing Tung / ACCT Course: UROP1100, Fall

Environmental, Social, and Governance (ESG) has become increasingly important in evaluating a company’s performance. Nevertheless, it remains unclear whether the investors in financial markets would take ESG into consideration. This project aims to understand whether certain types of investors have specific preferences to ESG by collecting data, manually matching datasets, and reading reports related to ESG from the proxy statement filed in https://www.sec.gov/edgar/searchedgar/companysearch.html (U.S. Securities and Exchange Commission). The targeted investors included mutual fund, hedge fund and endowments. With the limited number of proposals (1099) I have analyzed, no clear conclusions can be drawn. However, the data being worked on combined with the work of my teammates will help in identifying their respective preferences.

Institutional Investors and Environmental, Social, and Governance (ESG) Preferences Supervisor: YEGEN Eyub / FINA Student: WU Yifeng / QFIN Course: UROP1100, Fall

In this UROP report, articles related to ESG-topic are summarized for writer to understand the whole picture of this research project. Abbreviated for the words Environmental, Social, and Governance, ESG is a trending topic for Finance researchers to explore the externalities generation and elimination during company’s operation. This report summarizes key concepts of ESG development and debate, and gives out a recent example of ESG research that sheds light on potential research interests for UROP participants.

184

School of Business and Management Department of Finance

Hong Kong IPO Market Supervisor: ZHANG Chu / FINA Student: LAN Qing / QFIN Course: UROP1100, Spring

Were the existing shareholders really generous to offer a treat for new joiners? The answer was clearly no given its post-IPO stock price performance. Nevertheless, there are several interesting questions to explore. Why were retail investors so passionate about it back then? How could they avoid falling into the similar trap again? Are there any signals or red flags to watch for?

This case study examined the above questions from five aspects, namely pre-IPO dividend policy, change in senior management, the role of cornerstone investor in the IPO process, media and sell-side coverage, post- IPO financial performance. The study was based on information and data collected from prospectus, past interviews with the CEO done by media, discussions on online investment forums such as Xueqiu, news coverage, Baidu Index, and industry reports.

Hong Kong IPO Market Supervisor: ZHANG Chu / FINA Student: LEE Ho Ching / ECOF TAN Xinyue / FINA Course: UROP1100, Fall UROP1100, Fall

This study sought out to find whether there are individual variables that can explain the short-term and long-term stock returns regarding companies listed on the Main Board of Hong Kong Stock Exchange. This paper takes a sample of 1,046 companies that listed on the Main Board of Hong Kong Stock Exchange during 2002 to 2020, excluding firms that were listed by introduction, and runs predictive regression analysis between 10 independent variables and 12 dependent return variables. The results show that no individual variables explain the return patter of the stocks significantly well.

Hong Kong IPO Market Supervisor: ZHANG Chu / FINA Student: QIAO Yanyang / QFIN Course: UROP1100, Spring

In 2018, HKEX officially announced Chapter 18A, allowing biomedical companies without revenue or profit to be listed. The new policy affects IPO market, investment banks and boosted biotech industry substantially. By the end of 2021, 48 companies were listed under Chapter 18A, and more than 20 companies had filed the application forms. To better understand this kind of companies’ performance, stock price and its impact after listing, this report analyzes a typical biotech company-Brii Biosciences Limited, and focuses on its IPO process, post-IPO performance and impact.

185

School of Business and Management Department of Finance

Hong Kong IPO Market Supervisor: ZHANG Chu / FINA Student: SUN Beini / RMBI ZHAO Xuehan / QFIN Course: UROP2100, Fall UROP2100, Fall

This research consists of two parts. First, we calculated short-term and long-term returns of stocks listed in the main board of the Stock Exchange of Hong Kong together with both market-adjusted and sector-adjusted returns and examined the distribution of the returns. Second, we did prescriptive regression analyses of the expected short-term and long-term returns using best subset selection method and stepwise selection method based on information known upon IPO, such as firm-specific financial fundamental information, market condition and sector condition. And we drew three key findings about the stock performance prediction from the regression models from both short-term and long-term perspectives.

Hong Kong IPO Market Supervisor: ZHANG Chu / FINA Student: WONG Cheuk Wun / RMBI Course: UROP1100, Fall UROP2100, Spring

The Hong Kong IPO atmosphere remains active in 2021, recording 96 companies issuing their initial stocks. Nayuki Holdings Ltd, a tea-making company that IPO in 2021 is being studied in this paper. Despite the fact that the company had recorded a large pre-IPO subscription ratio, implying a massive demand for the company’s stock, its stock price recorded a significant amount of first-day drop as large as 15.76% upon its first day of listing. As of April 03, 2022, the accumulated drop has reached 72.15% from the offer price of HKD 19.8 to HKD 4.64. This serves as the motivation for this paper to perform a detailed investigation of the company’s background and business. After studying the global offering prospectus and related news which were published around the date of IPO, two major discrepancies have been discovered which might First, be the causes of the mentioned phenomenon.

Hong Kong IPO Market Supervisor: ZHANG Chu / FINA Student: WONG Yan Yin / QFIN Course: UROP1100, Spring

The rest of the paper will dive down into the company's valuation to explore the rationale behind the price movement of Kuaishou Technology. If it was underpriced, what is the incentive and reason behind it? If it as not underpriced, what is the cause of the unexpectedly high demand in the market? Does that indicate the group has chosen the perfect timing to get listed to maximize the profit? Finally, the paper will explore the reason behind the abnormal price movement of Kuaishou.

186

School of Business and Management Department of Finance

Hong Kong IPO Market Supervisor: ZHANG Chu / FINA Student: XU Han / RMBI Course: UROP1100, Spring

Right before its IPO on Hong Kong Main Board, Qingci is quite active in capital market. The biggest Stakeholder- G-bits potentially sold its stake to Alibaba Lingxi, Guangxi Tencent and Shanghai Hode. This is the first time that Alibaba, Bilibili and Tencent become cornerstone investors at the same time for one single games company. Qingci’s first quarter revenue in 2021 also largely increased. It seems that Qingci was fully prepared for its IPO. However, after its IPO with the subscription price of $11.2, its stock price keeps dropping until about one third of this original price. The report aims to study this unusual observation of Qingci Games and try to analyze the reason behind the decrease of stock price.

187

School of Business and Management Department of Information Systems, Business Statistics and Operations Management

Department of Information Systems, Business Statistics and Operations Management AI's Impact on the Financial Market Supervisor: CHEN Yanzhen / ISOM Student: WONG Yui Sing / IS ZHENG Bingyin / RMBI Course: UROP1100, Fall UROP1100, Fall

Since 2014, AI-assisted or Auto-generated news has become more important in mainstream media, and the amount of financial news reported, and the number of companies covered in news increased. Since media is one of the most important resources for individual traders to receive information, we looked into the impacts on financial markets after the rise of Robo-Journalism through this project and learn the methods of data manipulation at the same time. Our project can be divided into two steps. The first one is data collection and the second one is linear regression modelling. This semester, we completed the data collection part. We first downloaded articles from Factiva and then prof Chen helped grab the key information, which shows whether the articles were generated by AI or human beings. Secondly, we classified the articles into 3 categories: hybrid, robot, and human. Lastly, we compared the companies’ names collected from Factiva with the names used for merge. Next, we will build linear regression models based on the data collected.

Applicability of Deep Learning Solutions to Business Problems Supervisor: KWOK James Sai Ho / ISOM Student: JIA Tongyao / ELEC Course: UROP1100, Fall

This report aims to investigate the selection of specifications of convolutional neural networks (CNNs) for binary classification, including distinct types of input datasets and model structure hyperparameters. In this report, test accuracies are recorded and optimized by trying different parameters mainly based on the structure provided by keras_tuner on 14 different datasets. It is found that the number of layers, number of neurons, activation functions, and data spilt ratio will contribute to the final test accuracies. These results will provide a guideline for the CNN model building for binary classification problems.

188

School of Business and Management Department of Information Systems, Business Statistics and Operations Management

Who Becomes Venture Capitalists? Supervisor: KWON Ohchan / ISOM Student: CHEN Ze Yu / ECOF LI Zixuan / FINA YANG Xixuan / OM YE Zijian / RMBI Course: UROP1000, Summer UROP1000, Summer UROP1000, Summer UROP1000, Summer

When it comes to discovering and encouraging high-growth companies, venture capital (VC) is essential. However, the sector has come under fire for being overly controlled by white men, at least in the US. Imbalanced financing opportunities may result from the biased composition of venture capitalists. Despite the significance, we lack scientific knowledge of how and who becomes a venture investor.

There is a claim that a gender disparity in entrepreneurship funding is a direct result of the gender disparity in early-stage investment. In this research project, we examine two associated questions to support this argument: (1) what variables affect the entry of highly skilled men and women into the venture capital industry, and (2) does knowing other female venture capitalists boost a female's probability of starting her own business. By constructing a comprehensive career history of MBA graduates from the Harvard Business School between 2000 and 2015, we would like to perform empirical investigations. This paper provides examples of the data collecting methodology, examines the current state of play, and outlines subsequent phases of.

Who Becomes Venture Capitalists? Supervisor: KWON Ohchan / ISOM Student: DENG Xinyi / RMBI LAM An Yin Andrew / ECOF Course: UROP1100, Fall UROP1100, Fall

There is an argument that gender imbalance in early-stage funding leads to gender imbalance in entrepreneurship. To inform this argument, in this research project we examine two related questions: (1) what factors determine highly qualified male and female professionals’ entry into the venture capital industry, and (2) whether having a social network with same-gender venture capitalists increase a female’s likelihood of pursuing entrepreneurship. We aim to conduct empirical studies by constructing detailed career history of MBA graduates from the Harvard Business School between 2000 and 2015. This report illustrates the data collection strategy, discusses the progress so far, and details the next steps.

189

School of Business and Management Department of Information Systems, Business Statistics and Operations Management

Gender-biased Perception towards Entrepreneurship Supervisor: KWON Ohchan / ISOM Student: LEE Ka Yi / RMBI QUACH Tuan Kiet / SBM Course: UROP1100, Fall UROP1100, Fall

Compared to male entrepreneurs, female entrepreneurs need to overcome biased evaluation. In this research, we examine if female entrepreneurs frame their entrepreneurial motivation differently than male entrepreneurs, and whether such differences in framing strategy contributes to reducing the bias. We examine our research question in the context of Shark Tank, a popular reality TV show in the United States. Our initial results suggest that female entrepreneurs emphasize their user experience more frequently than male entrepreneurs, even within the same industry. To understand the impact, we plan to examine the implications for fundraising by employing micro-level data and conduct text analysis to see how framing as user-entrepreneur shapes the follow-up questions from investors.

Gender-biased Perception towards Entrepreneurship Supervisor: KWON Ohchan / ISOM Student: WONG Tsz Kwan / IS/GBUS Course: UROP2100, Fall UROP3100, Spring

Gender homophily in start-up fundraising is often cited as an important factor contributing to the gender gap in entrepreneurship. Because existing studies focus on the investor side explanations, our understandings of why we observe gender homophily in start-up funding is incomplete. To fill this important gap, we focus on the entrepreneur side and examine whether entrepreneurs prefer investors of the same gender and the value attached to such homophily. We measure entrepreneurs’ preference for gender homophily with investors using data from Shark Tank. We analyze entrepreneurs who have received multiple offers from investors of different genders to control heterogeneity in entrepreneurial quality. Our findings contribute to entrepreneurship and homophily literature and provide practical insights in reducing the gender gap in entrepreneurship.

190

School of Business and Management Department of Information Systems, Business Statistics and Operations Management

Gender-biased Perception towards Entrepreneurship Supervisor: KWON Ohchan / ISOM Student: XIA Xiaotong / MAEC Course: UROP1100, Fall

Unemployment insurance (UI) provides cash stipends to unemployed workers who actively seek employment and is an essential part of social security system. However, is longer duration of UI always better for the society? What’s the underlying relationship between average working efficiency and UI? Under the supervision of Prof. KWON Ohchan, I received a great insight into this topic through readings, discussions, data findings, and analysis, after which I got had a better understanding of how UI affects the society. This report consists of mainly six parts, including introduction, objectives, findings & analysis, individual reflection, reference and appendix.

A Study on Amazon Aggregators Supervisor: KWON Ohchan / ISOM Student: DENG Hanwen / WBB Course: UROP1100, Spring UROP2100, Summer

Scholar have examined the impact of buyouts in different sectors. An increasing number of buyouts target online marketplace sellers recently, but we do not know how these buyouts affect these sellers in online marketplace settings. In this paper, we document changes in prices, sales rank, customer satisfaction, product introduction and discontinuation, and other characteristics of Amazon sellers experiencing buyouts. We find that operational practices improve after private equity buyout, as restaurants become cleaner, safer, and better maintained. These changes are particularly apparent when private equity partners have prior industry experience (maybe experience matters?). Our results suggest that by bringing in industry expertise, private equity firms improve firm operations.

Who to Blame the Ricky Gig Economy: Jockey vs. Horse? Supervisor: KWON Ohchan / ISOM Student: LEE Chaewon / SBM Course: UROP1100, Summer

Gig economy is important. However, there is a growing concern that gig economy platforms may incur unsafe practice. A change in fee structure implemented by a ride-hailing platform, where drivers’ fees were changed from an upfront charge for a fixed amount of time on the platform to a per ride commission. The upfront participation fee leads to a less flexible arrangement, which is likely important for part-time workers. In this paper, I propose an evidence to understand whether it is because of inexperienced gig workers, or because of the issues related to business models. We will empirically examine one Korean platform.

191

School of Business and Management Department of Information Systems, Business Statistics and Operations Management

Identifying Entrepreneurs' Storytelling Strategies Using Twitter Data Supervisor: KWON Ohchan / ISOM Student: HO Ho Yi Alexis / QFIN JIANG Yuxuan / ECOF Course: UROP1100, Summer UROP1100, Summer

An accurate prediction of whether an investor will proceed to a capital venture for small businesses is indispensable to an effective classification of Twitter messages. However, as the preliminary data cleaning process took longer than we expected, we were unable to verify the selected model’s accuracy using data from reality. Rather, we will present this cookbook to illustrate our methodologies using analog data from the R package. After extensive literature reviews, we identified Reisenbichler and Reutterer’s work in topic modeling in marketing in 2018 as our major discussion base. Along with our analysis, we supplemented the necessary information as instructions that needed to be known by our successor, the Research Assistants of this following research program series.

Digital Leadership and Digital Transformation Supervisor: LEE Dongwon / ISOM Student: HUI Ka Yee / FINA Course: UROP1100, Spring

There is a wide consensus on the prominence of leadership for digital transformation, however gap between modern leadership and digital leadership is still something researchable as synchronization of companies’ broad direction towards digitalization and employees’ digital adaption remains a difficulty face by the leaders. To support the process of digital transformation, how leaders can better act as linking bridge between mission and humankinds by using innovative strategies that differ from modern leadership style will be the most important direction of this report. The purpose of this report is to discover similarities and differences between traditional leadership and digital leadership by looking at their practice and efficacy under various circumstances to advice a detailed guideline to help leaders better understand digital transformation from human perspectives.

On Identifying Web3 Business Opportunities and Challenges Supervisor: LEE Dongwon / ISOM Student: CHEN Yuewen / ACCT Course: UROP1100, Spring

With the development of the 21st century, the world has ushered in a new trend in the field of high technology, which is the advent of Web3.0 era. This REPORT focuses on the theme of Web3.0, and introduces the opportunities and challenges in Web3.0 with two major themes, DAO and NFT. The report is divided into four main parts, the first part is a brief overview of the background information of Web3.0. The second part is about NFT, its features, its application in the market and the development trend in China. The third part introduces the features of DAO, the good and bad and the trend in China. The last part is some selfreflection in the process of doing research by myself.

192

School of Business and Management Department of Information Systems, Business Statistics and Operations Management

On Identifying Web3 Business Opportunities and Challenges Supervisor: LEE Dongwon / ISOM Student: KUAN Ho Man / SBM Course: UROP1100, Summer

Non-fungible tokens (NFT) are indisputably one of the most absurd asset classes that break out in 2021-2022. In short, NFT is a form of digital asset leveraging the properties of smart contracts and blockchain, allowing them to distinguish themselves from each other, therefore non-fungible. NFTs can be tied to physical assets as a certification of ownership, yet it is not hard to observe that in the past year, the increasing activeness of the market does not have much to do with physical assets but with virtual arts and game-related collectibles, things that traditionally don’t hold much tangible value. In this paper, I hope to provide and analyze different events that occurred in the past few years and provide various metrics that can help evaluate how and whether these treatments affect the nature of different NFT projects and Web3 infrastructures.

On Identifying Web3 Business Opportunities and Challenges Supervisor: LEE Dongwon / ISOM Student: ZHOU Xiaoyu / RMBI Course: UROP1000, Summer

With the development of digital assets, the realistic importance of digital asset pricing on financial reporting, taxation, and investment considerations has grown with time. In this paper, we identify the current digital asset pricing models and compare them with traditional asset pricing models and startup valuation models. We also discuss the factors driving the price of digital assets and the dynamic valuation factors yet to be considered in pricing models. There is a migration of the idea of traditional valuation methods on digital asset pricing based on the security or commodity nature of the assets, as well as completely new methods or factors revealing the value of digital assets with utility nature. Apart from the traditionally considered factors, there are also digital-related factors that affect digital asset pricing and crypto factors to consider in valuation.

Deep Learning NLP in finance Supervisor: YANG Yi / ISOM Student: HU Chenxi / COGBM Course: UROP1100, Summer

Sentiment analysis has been widely adopted in finance fields as it facilitates price predictions and risk management in quantitative measures. The introduction of deep learning models such as BERT has also increased the accuracy thus boosting the performance in various tasks such as stock price predictions. However, deep learning models inherently suffered from data shifts and spurious correlations, which are prevalent in the stock market. In this paper, we present how spurious correlations and data shifts hinder the performance of the deep learning models on sentiment analysis in stock markets and demonstrate how we may mitigate this.

193

School of Business and Management Department of Information Systems, Business Statistics and Operations Management 194 Deep Learning NLP in finance Supervisor: YANG Yi / ISOM Student: WANG Xiaopeng / DSCT Course: UROP1100, Summer

Existing NLP systems rarely separate numbers and words in the text when embedding them. This obviously contrasts with the consensus in neuroscience that the human’s brains treat numbers differently from words. For instance, most NLP models are not specially designed for processing numbers, most of which either discard numbers directly or replace them with UNK tokens. However, given the ubiquity of numbers especially in the financial area, it would be beneficial for us to enable NLP models to understand numbers more effectively.

Deep Learning NLP in finance Supervisor: YANG Yi / ISOM Student: ZHOU Zhuorui / COGBM Course: UROP1100, Spring

Earnings conference calls are the most interactive ways between companies and investor. To objectively evaluate the performance of different business components for a company, it is a common practice to categorize different sectors within analyzed company’s business. While the overall objective of this research projects aims to automatically categorize business topics for given company and analyze the sentiment of company’s performance accordingly, this report mostly comprises of exploratory analysis and serves as preparation for the overall project. The report focuses on models, such as LDA, InfoMap, experimented for the “topic selection” part. Latest development in topic models is also introduced in this report.

194

School of Business and Management Department of Marketing

Department of Marketing E-commerce and the COVID-19 Epidemic Supervisor: HELSEN Kristiaan / MARK Student: YU Chin Ting / FINA Course: UROP1100, Fall

This report aims to study the ecommerce trend in Hong Kong after the epidemic, particularly in the senior segment. The effect of specific marketing campaigns and the existence of O2O stores are both tested out in the following. This study used HKTVmall as an example.

195

School of Business and Management Department of Management

Department of Management Multiculturalism, Social Judgments and Decisions Supervisor: CHAO Melody Man Chi / MGMT Student: KIM Hayeon / CBGBM Course: UROP4100, Fall

In modern society, mental health issues are greatly arising, but social stigma and high costs are hindering people from addressing the issue. This UROP project investigates a therapeutic writing where people would write honestly about their feelings to overcome emotional burdens, as an easily accessible, practical and affordable way to maintain good mental health. As the mental well-being encompasses many different aspects, this research specifically focuses on burnout and emotional experiences. The intervention study was conducted over a week with five different time points of pre-survey, three writing interventions and postsurvey. Whether the participant’s well-being conditions have improved or not after the interventions have been evaluated using statistical analysis.

Multiculturalism, Social Judgments and Decisions Supervisor: CHAO Melody Man Chi / MGMT Student: NG Ying Lam / ECOF Course: UROP1100, Summer

Stereotypes often portray old adults as ill-tempered and easily upset. However, empirical and neurological studies have shown the opposite, that the elderly are more capable than younger adults in controlling their emotions, while increase in age tends to change emotional expression tendencies positively. The biased view on emotional expression of the elderly may have its root in the facial features of old adults, such as wrinkles and folds, which makes the emotion of the elderly harder to be interpreted accurately. Tendencies of the elderly experiencing relatively fewer intense emotional experiences may also create a misimpression that the elderly lack empathy. The existence of emotional-related stereotypes over old adults might create hindrances for the elderly in job-seeking and daily communication with others.

Social Perceptions, Justice, and Diversity Supervisor: CHAO Melody Man Chi / MGMT Student: CHEN Jingruo / QSA Course: UROP3100, Fall UROP4100, Spring

Most literature on ideal worker norms has suggested a general organizational expectation that an ideal worker would be an individual wholly dedicated to work and unencumbered from family or personal life. We aim to use the intersubjective perception approach to explore the whether and how individuals endorse the ideal worker norms as their personal values across generations, work positions and cultural backgrounds; (2) their intersubjective perceptions of the ideal worker norms; (3) examine its association with employee’s psychological well-being and work-related behavioral patterns.

196

School of Business and Management Department of Management

Social Perceptions, Justice, and Diversity Supervisor: CHAO Melody Man Chi / MGMT Student: JHA Spriha / MGMT Course: UROP1100, Fall

Emojis have become an important part of those who participate in computer-mediated communication these days. There is a clear disagreement regarding whether it is appropriate to use them in workplace communication. For some, it is a natural and normal thing to do in order to communicate non-verbal cues missing from computer-mediated communication, while others are afraid of coming off as less competent. The current research aims to examine whether these beliefs about using emojis have any real-life implications and if these beliefs are affected by a person’s position in an organization. How does a sender’s status influence the effect of his/her use of emoji on his/her perception? What is the downstream consequence of the effect of emoji on perceptions? While existing literature has inconsistent findings about perceptions of warmth and competence, the current study (N=217) found that emojis communicate warmth, with no effect on perceived competence, ultimately increasing the willingness to work. The communicated warmth is not affected by the person’s hierarchical status, while perceived competence solely depends on the status. Implications of the findings are discussed.

Social Perceptions, Justice, and Diversity Supervisor: CHAO Melody Man Chi / MGMT Student: KONG Tsz Yan / ACCT Course: UROP1100, Spring UROP1000, Summer

Existing literatures describe ideal worker as a worker with high level of devotion, productivity, and obedience. While literatures have argued that ideal worker norm as a universal norm, there is some evidence proving this norm is not necessarily endorsed and embodied by everyone. Research has found that employees experience conflict between the ideal worker norm and their professional identities (Reid, 2015). Some organizations even implement systems that discourage a culture that praise the ideal worker norm (Kelly et al., 2010). This report examines how the conflicting perception between individuals’ personal value on ideal worker and the ideal worker norm perceived in society affect one’s psychological wellbeing.

197

Abstracts of UROP Projects 2021-22 School of Humanities and Social Science

School of Humanities and Social Science Division of Humanities

Division of Humanities Korean War Prisoners Oral History Interview Transcription Supervisor: CHANG David Cheng / HUMA Student: XUE Yuyuan / GCS Course: UROP2100, Fall

As an essential part of the Second World War, the Burma Campaign was a battlefield where the Chinese, British and American troops worked together as alliance. Faced with the prospect of fighting together with different languages and weapons, military translators and interpreters were bound to be an important part of the cooperation that the three sides were able to develop smoothly. Where and how many Allied interpreters from China were recruited, their role in the Burma campaign, the difficulties they faced, and their personal experiences are all questions that Professor Chang David Cheng's current research project is intended to explore. Using the documentation provided by Professor Chang as a starting point, I have used primary and secondary sources to collate information on Allied interpreters. This report includes an introduction to the project, my work content and methodology, as well as my findings and personal reflections at this stage.

History of Prisons, Penal Colonies and Quarantine Centers in Asia Supervisor: SMITH Jenny Leigh / HUMA Student: DONG Bochen / GCS Course: UROP1100, Fall

In Fall 2021, I did some research into the history of leprosy in East Asia under the supervision of Professor Smith. First, I investigated Hei Ling Chau, a former leper colony in Hong Kong. Second, I examined the Tai- Kam Island in Guangdong. Third, I compiled a chart for some leprosy institutions in Taiwan. Fourth, I was advised to read journal articles about leprosy in Asia-Pacific, especially in Australia, the Philippines, and Hawaii from the 1920s to the 1970s. Besides, I gave a presentation on the history of leprosy in Japan. I paid attention to the maltreatment and slaughter of Guangdong lepers in the 1930s, too. By comparing the ways lepers were treated in different regions and times, I made some generalizations and gained deeper understanding of social progress. Overall, the project provided a starting point for my exploration of the history of leprosy.

200

School of Humanities and Social Science Division of Social Science

Division of Social Science AI for Social Innovation: Leveraging Chatbots to Combat Sexual Harassment Supervisor: GIETEL-BASTEN Stuart Arthur / SOSC Student: CHAN Wan Ting / ECOF GANGAR Simar / COGBM LEUNG Tsz Ching / CPEG LI Wanshan / IS MOK Nga Yiu / COGBM NG Lok Lam / ELEC ZHU Qiyun / GCS-IHSS Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer UROP1100, Spring UROP2100, Summer UROP1100, Spring UROP2100, Summer UROP1100, Fall UROP2100, Spring UROP3100, Summer UROP1100, Spring UROP2100, Summer UROP1100, Fall UROP2100, Spring UROP3100, Summer UROP1000, Summer

The project is inspired by many news reporting sexual harassment cases in university orientation camps. By building the chatbot, we hope to support the victims and raise awareness of sexual harassment. The chatbot consists of three major functions, i.e., chitchat, education, and documentation. Currently, we have successfully implemented the flowcharts of conversations for all 3 functions. Apart from that, we are working on improving the flexibility of the content by using computer language C# and enhancing the user experience by researching the existing chatbots in the market. The chatbot is anticipated to be launched by next academic year on school websites and some student mobile apps.

201

School of Humanities and Social Science Division of Social Science

AI for Social Innovation: Leveraging Chatbots to Combat Sexual Harassment Supervisor: GIETEL-BASTEN Stuart Arthur / SOSC Student: TSANG Chin Yung Virginia / COMP/CS Course: UROP1100, Fall UROP2100, Spring

With an understanding of the landscapes of AI and social innovation from the previous UROP course series, this semester we worked on exploring technical capacities in AI and aspects of programming. We started by deciding on chatbots as a means to leverage sexual harassment issues, then designing the content and conversation flow of the conversation agent from literature reviews and systematic research in the past. At the same time, we compared various vendors on chatbot frameworks in terms of technical capacities and financials. With the functionalities and user friendliness that the Microsoft Bot Composer Framework offered, it best supported our two main features of the chatbot: Documentation of sexual harassment events and Education. By the end of the project, we have created the Education section of the chatbot on the Composer, which includes different sub-conversations (sub-branches), and an additional chit-chat function that makes the chatbot response natural.

Reconstructing Historical Lives from Demographic Data Supervisor: GIETEL-BASTEN Stuart Arthur / SOSC Student: JIAN Yu Kei / SENG Course: UROP1000, Summer

In UROP1000, my job is data transcription, which is to transcript data of the Askham Bryan baptismal registers, the York St. Crux baptismal registers, and correct the transcription of the Linton-in-Craven marriage registers. Data entry involves carefully selecting key information from registers and entering it into excel tables. During this project period, I learned how the raw data looks like and how parishes in late 18th to early 19 th century registered the baptisms and marriages. I also learned methods of cleaning and processing raw data. Additionally, I conduct a simple study to find the relationship between gender, location and population movement. The followings are my reflections and findings.

202

School of Humanities and Social Science Division of Social Science

Reconstructing Historical Lives from Demographic Data Supervisor: GIETEL-BASTEN Stuart Arthur / SOSC Student: LAM Ka Yan / GCS Course: UROP1100, Fall

In this project, I worked on the historical marriage records of Ghana since these historical images contain population data that are extremely helpful in helping researchers to understand not only the past but also the present. I have registered for a free account at FamilySearch and access the marriage data there. Throughout the research process, our team focused on 2,694 images of marriage certificates in the time period “1972-1976 (vol 1, RG18-1-10)”. We mainly transcribe the year of marriage, the occupations of the groom and the bride, the occupations of the father of the groom, the occupations of the father of the bride, and their signature or mark indicated next to their signature. Last but not least, as the marriage certificate may be scanned more than one, we need to avoid duplicating the same marriage twice.

Reconstructing Historical Lives from Demographic Data Supervisor: GIETEL-BASTEN Stuart Arthur / SOSC Student: LAM Oi Tung / GCS-IHSS Course: UROP1000, Summer

In the 2021-22 Summer semester, I have enrolled in the UROP1000 - Reconstructing Historical Lives from Demographic Data and work under the supervision of Professor Stuart GIETEL-BASTEN and his research assistant Rachael Cheung. During the past two months, I am responsible for dealing with different data and have more in-depth understanding of how research is conducted and processed, and also develop certain skills for analyzing historical documents. In this report, I would categorize my experience of working as a research assistant intern into three parts, which are (1) the job content of the program, (2) reflection on the work and (3) the benefits and potential outcome of joining the course.

Reconstructing Historical Lives from Demographic Data Supervisor: GIETEL-BASTEN Stuart Arthur / SOSC Student: LIU Junde / QSA Course: UROP1100, Fall

During the course of the summer of 2021, my teammates and I built two sets of data holding the birth and death records of the State of Delaware from the year 1890 to 1920. My report for UROP 1000 this summer attempted to link the birth and death records from the two data, but to my demay, I only managed to achieve a match rate of 5% after some considerable efforts which involved the use of cloud computing. For this UROP 1100 report, I decided to take a step back and examine the data by gathering some descriptive statistics. It can not only help me to better understand the data at hand, but also potentially provide me with some insights as to the reasons for my failure in summer.

203

School of Humanities and Social Science Division of Social Science

Reconstructing Historical Lives from Demographic Data Supervisor: GIETEL-BASTEN Stuart Arthur / SOSC Student: NGAU Ming De Pearl / GCS Course: UROP1100, Fall

Focusing on the links between population and policy in Asia, “Reconstructing Historical Lives from Demographic Data” has adopted archival research method gathering historical data from FamilySearch to analyse critical social issues among Asian countries, including the emergence of low fertility and the consequences of this in terms of population ageing and growth. Based on the extensive empirical research done in Asia, archival-based analytical research is thought to develop a more thorough knowledge of social concerns in Asia and revise and support the implementation of policies.

Repressing the Emotion: Censorship of Popular Culture under Authoritarianism Supervisor: HONG Jean (Ji Yeon) / SOSC Student: CHAN Ho Wai / GCS Course: UROP1100, Fall

The political climate in Hong Kong encountered a drastic change after 2019. Major events like the anti-extradition law movement in 2019, legislation of National Security Law 2020 and the reformation of the electoral system of the legislative council created huge changes in Hong Kong’s politics. With a series of important social movements and institutional reforms, it is unavoidably that the general public and the press changed their attitude towards the HKSAR government and the central government. The paper demonstrates the changes in public sentiment as well as the changes in magnitude of the propaganda effect. By adopting textual analysis through R studio, sentiment analysis and topic modeling would visualize the results. It was discovered that the sentiment score towards the HKSAR government had improved from 2019- 2021. It was also discovered that the authority may make tremendous efforts in propaganda where the propaganda wordings appeared more frequently. Furthermore, the topic of the NewsPaper articles primarily focussed on the Social event in Hong Kong. For example, the social protest in 2019, national security law legislation in 2020. The research findings would provide insights for future analysis and further interpretation related to the relationship between the government and the media.

204

School of Humanities and Social Science Division of Social Science

Political Astroturfing Worldwide: How to Manipulate Public Opinion on Twitter Supervisor: KELLER Franziska / SOSC Student: LUI Hiu Lum / ACCT Course: UROP1100, Fall

This report aims to investigate the characteristics of Twitter accounts that may be participating in political astroturfing. Target accounts are accounts that followed multiple Russian official accounts. Account information is being examined, for example, creation date and profile details. Two account creation spikes have been identified in 2017 and 2020. Further analysis is conducted on these 2 groups of accounts.

Both groups demonstrate similar behavior – spike groups have more inactive accounts with less active accounts, smaller proportion of private accounts and larger proportion of accounts that have never tweeted, comparing to their respective control groups. However, the 2017 spike group has shown statistically significant difference in mean follower number, while the 2020 spike group does not.

Political Astroturfing Worldwide: How to Manipulate Public Opinion on Twitter Supervisor: KELLER Franziska / SOSC Student: MOON Kyoungin / QSA Course: UROP2100, Fall

This report examines the Twitter accounts that are affiliated with South Korean diplomatic accounts in an attempt to trace any suspicious pattern that possibly indicates the presence of political astroturfing. Oftentimes, politicians use astroturfing accounts to elevate their status or to influence public opinion on social media (Keller et al., 2020). This report is divided into two sections: 1) overall analysis of the South Korean-affiliated accounts and 2) comparative analysis with follower accounts that are affiliated with Chinese diplomatic accounts. The first section finds no suspicious patterns among accounts affiliated with South Korean diplomatic accounts. By comparing South Korean-affiliated accounts with Chinese-affiliated accounts, this report highlights several patterns, including accounts’ profile and activity status, that help to examine how the behavior of astroturfing accounts differ from that of regular accounts.

205

School of Humanities and Social Science Division of Social Science

The Covid-19 Pandemic: Social Science Perspectives Supervisor: SAUTMAN Barry / SOSC Student: AU-YEUNG Lok Ching / GCS Course: UROP1100, Fall

The widespread of Coronavirus disease (COVID-19) since late 2019 has led to lots of confirmed cases and death around the world, which give rise to the invention of vaccines by various countries. This paper explores the major vaccine production countries, characteristics of different types of vaccines and discusses the criticism of vaccine diplomacy by analyzing the discourses made by various politicians. Vaccine’s donation and sales patterns of countries (mainly the United States and China) would be used to illustrate the discourses that have been made to blame China as the country to evoke and engage vaccine diplomacy have no evidence and facts support.

The Covid-19 Pandemic: Social Science Perspectives Supervisor: SAUTMAN Barry / SOSC Student: MA Sum Yin Tracy / GCS Course: UROP1100, Fall

The coronavirus disease (COVID-19) has been affecting the world for almost 2 years. China, the United States, and other emerging economies have performed impressively in emergency vaccine research and development. However, the vaccination situation in different countries varies greatly. There is a serious imbalance and differentiation in the international distribution of vaccines. From the aspects of research and development, production, application, effects, and prospects of the new vaccine, the vaccine diplomacy argument is the latest manifestation of the politicization of the anti-epidemic issue. Which countries criticize one another for using the vaccine as a political tool to gain a greater influence in the globe. In particular, it often happens in the US towards China. From the findings collected, we will analyze if the criticisms are reliable and legitimate.

206

School of Humanities and Social Science Division of Social Science

The Covid-19 Pandemic: Social Science Perspectives Supervisor: SAUTMAN Barry / SOSC Student: OU YANG Cai Waverly Chua / GCS Course: UROP1100, Fall

The donation or the sale of the Chinese vaccine caught global attention as some assumed that China might gain political influence by doing so. Based on the definition from scholars, we define 'vaccine diplomacy' contain four elements, including trust enhancement; global health leadership; good quality of 'made-in- China' products; and China as a technologically advanced nation. We investigated how Indonesian and Uzbekistan report on the Chinese vaccine and assessed whether their reports coincides with 'vaccine diplomacy. We concluded that Chinese vaccine diplomacy might not be as effective in increasing China's influence as scholars considered as the media did not or could not draw a connection between the Chinese vaccine and the four elements mentioned above.

China/Africa Links Project Supervisor: SAUTMAN Barry / SOSC Student: CHAN Hoi Ho / RMBI Course: UROP1000, Summer

This report forms part of a larger China / Africa Links Project and investigates the agricultural partnerships between China and Africa. In particular, the Democratic Republic of Congo and Burundi has been selected as countries of investigation for their varying degrees of economic relationship with China. While the Democratic Republic of Congo is one of the largest trading partners in terms of natural resources with China, Burundi is one of the most impoverished countries in the world and relies on China for aid. Special attention has been given to the economic relations and agricultural sectors of these two countries before proceeding to a more thorough discussion of agricultural partnership primarily through Agricultural Technology Demonstration Centers located in both countries. Towards the end of the report, using the basis of the indiscriminate aid and partnership China provides African countries, this report further argues that the Chinese-African agricultural technology is a shining example of altruistic South-South Cooperation.

207

School of Humanities and Social Science Division of Social Science

China/Africa Links Project Supervisor: SAUTMAN Barry / SOSC Student: CHENG Yu / GCS Course: UROP2100, Fall

This progress report is part of HKUST's Undergraduate Research Opportunity Program and is targeted on the China- Oceania Links research, under the guidance and supervision of Professor Barry Sautman and Professor Yan Hairong. This report aims to summarise the contribution to the project over the semester, with an emphasis on the political and economic nature of China's relations with countries around the world, concentrating on the trade and investment connections among the bilateral relationships, the agricultural cooperation as another focus, and the evaluation of such connections by international public opinion. This progress report focuses on the relationship between China and the countries of Oceania. The research methodology is mainly a documentary review and data collection, drawing on data from the World Bank, WTO and specific governments to obtain an overview of the political and economic status in which each Oceania countries, measuring the degree of economic dependence between China and that specific Oceania country, referring to official treaties, documents and press reports to ascertain different patterns of agricultural cooperation between China and Oceania countries, and finally collecting a wide range of data to suggest the degree of Chinese land investment or land holding in Oceania countries.

China/Africa Links Project Supervisor: SAUTMAN Barry / SOSC Student: QIAO Tianyi / GCS Course: UROP1100, Spring

Under the project, I worked to gather information related to China-Africa agrobusiness links and drafted a full-length country report on South Africa. I made flexible use of both Chinese and English sources to investigate the nature of cooperation between countries, seeking to account for the cooperation progress chronologically. I also launched primary interviews to related bodies to get first-hand information. In addition, I compiled data made available by factiva to record and trace localization process of Chinese companies and projects in African countries for future references on investigating the diagram between China and Africa. Valuable lessons on how to gather data, conduct individual research, as well as sort information for use were obtained.

208

School of Humanities and Social Science Division of Social Science

China/Africa Links Project Supervisor: SAUTMAN Barry / SOSC Student: YIU Wing Kiu Hayley / GCS-SOSC Course: UROP1000, Summer

China’s agricultural aid to Africa was initiated at the end of the 1950s. Over the past decade, the market-driven agricultural technology transfer model has been applied by China to African agricultural assistance. On the other hand, due to various reasons such as insufficient agricultural infrastructure, lack of breeding capacity, and poor level of agricultural mechanization, the agricultural development and food production in Africa have been severely restricted. To this end, the Chinese government has cooperated with different organizations, such as Food and Agriculture Organization of the United Nations (FAO), to help solve the food security crisis in African countries through food aid and other means. This paper aims to find out to what extend China has developed programs or participated in terms of designing agricultural skills for African countries. This paper focuses on the experience of FAO to find out what Chinese organizations or individuals have done in terms of transferring skills and training program.

Understanding Experiences in Historical Environments Supervisor: TAM Kevin Kim-Pong / SOSC Student: CUI Boyi / EVMT Course: UROP1000, Summer

This is a progress report of UROP 1000: Understanding Experiences in Historical Environments. Historical environments are meaningful for human society. Their maintenance and development rely on visitors. Return visitors are important because they revisit the sites and spread positive word-of-mouth. Revisit intention has been deeply discussed though some inconsistency persists. Revisit experiences have not attracted much attention. So far, we have collected studies of revisit intention and summarized them in a document. A questionnaire has been designed for both questions. The second half of this progress report is a reflection of my learning process. Extensive reading, reflection, effective communication, and software skills help me continue doing the project.

209

School of Humanities and Social Science Division of Social Science

Understanding Experiences in Historical Environments Supervisor: TAM Kevin Kim-Pong / SOSC Student: PENG Wanwen / GCS Course: UROP1000, Summer

This is a progress summary of what I did in this UROP 1000: understanding experiences in historical environments. Several specific definitions of concepts are given at the beginning. Then origins of the three stages are summarized. The project starts from the Pre-Insitu-Post Three-step model, develops two specific directions, and finally scopes down to revisit intentions and experiences, including personal changes during this process. Knowledge of those three stages is introduced. The plan for this project is briefly described. The final part, reflection, describes the research method we applied and the skills I got. For lacking existing literature studying revisit experiences, the questionnaire we designed for this topic is attached to the appendix.

Environmental Blessing or Curse: Emission Effects of the Belt and Road Energy Infrastructure Projects in Africa

Supervisor: WANG Jin / SOSC Co-supervisor: LIN Yatang / SOSC Student: LEUNG Ho Fung / ECOF Course: UROP1100, Spring

This UROP project focuses on the Belt and Road Initiative and the environmental impacts it has on Asia, Europe and Africa countries. This report concluded the research and findings conducted in the Spring of 2022, mostly on data cleaning and matching. Python was used to clean and match data from Turkey, Ukraine, Chad, Sudan, Rwanda, Kyrgyzstan, Djibouti, Kazakhstan, Costa Rica, Egypt, Venezuela, Panama, Brunei, Fiji, Peru, United Arab Emirates, Korea, Maldives, Mali. This research aims to provide policy implications to international development policy design with a focus on energy infrastructure.

210

School of Humanities and Social Science Division of Social Science

Evaluate Economic Impacts of Metro Rail in Hong Kong Supervisor: WANG Wen / SOSC Student: NG Kai Yuen Dominic / QSA Course: UROP1100, Fall

High-quality public transport and public services attract more people to the railway stations and arouse appreciation in residential and commercial properties. Housing owners and local businesses gain from such changes while housing renters lose. The socio-demographic changes and the labor market changes can further redistribute the welfare among different residents. We combine the data of GHS and HK_POI with demographic characteristics and data about the areas around stations, such as housing, shopping malls, and offices, to analyze the demographic and distributional changes. The analyses are indispensable for policy implications such as the operation of a metro rail system, the provision of public housing, and the consideration of social equality and efficiency.

Evaluate Economic Impacts of Metro Rail in Hong Kong Supervisor: WANG Wen / SOSC Student: WU Kwan Yu / QSA Course: UROP1100, Fall

With the rapid population growth and urbanization, big cities such as New York, Chicago, Beijing, and Shanghai are developing and expanding public transit systems, especially the metro rail system. Most of these public infrastructures incur huge deficits. However, Hong Kong’s MTR Corporation (MTRC) has been highly profitable. MTR has expanded its business by constructing new lines. Until 2021, the total route length of MTR’s railway system is 197.1 km compared to 175 km in 2014. MTRC earned the most profit from the properties surrounding the metro rail stations instead of the operations of the metro rail lines. This is a business model promoted and implemented in Hong Kong ---“Rail plus Property” (R+P) model ---in which the government grants MTRC with land development rights along the route.

211

School of Humanities and Social Science Division of Social Science

The Impacts of Green Finance Supervisor: WANG Wen / SOSC Student: XIE Xiaoming / QSA Course: UROP1100, Fall

This project investigates the impacts of environmental regulations on reducing pollution through encouraging green innovation. The two regulations investigated are the Green Credit Policy and the pollution tax system. The main tasks for this program include a literature review and data exploration. According to a literature search, some environmental regulations reduce pollution while others have reverse effects, causing firms to produce more to offset the additional pollution cost. Green technology may give rise to cleaner production and reduce pollution costs per production, allowing firms to produce more and emit more pollution. Data exploration shows a positive correlation between policies and green technology and a positive correlation between technology and pollution.

The Impacts of Green Finance Supervisor: WANG Wen / SOSC Student: YAO Wing Lam / QSA Course: UROP1100, Fall

The government policy can always influent firms’ decisions and behaviors. In March 2010, the former U.S. President Barack Obama established the Affordable Care Act (ACA), which requires firms to offer some affordable health care insurances to employees. However, this may increase the operating costs of firms. Therefore, the companies may decide to take actions to reduce the cost of health care insurance such as switching to cleaner production, moving to locations with a better environment or cheap labor. In this project, we investigate if ACA has influenced US companies’ decisions in production, location, and employment. In addition, we will explore whether the Act can increase cleaner productions and decrease pollution emissions.

212

School of Humanities and Social Science Division of Social Science

Ridership and Public Transit Expansion Supervisor: WANG Wen / SOSC Student: WANG Yiwei / QSA Course: UROP1100, Summer

This report is a progress report of the project on ridership and public transit expansion. Specifically, this project investigates the effects of public transit expansion in the U.S. and how it may affect ridership and commuting patterns of residents. The main content of this semester includes data downloading and data cleaning. The source of the data includes Census Transportation Planning Products Program (CTPP) and American Community Survey (ACS) 5-year estimates. I finished downloading and cleaning the tract-level transportation data and census data using Stata in UROP. This report will review the process of downloading and cleaning data.

Identifying Protest Events in China with Social Media Data Supervisor: ZHANG Han / SOSC Student: FANG Zihang / GCS Course: UROP1100, Fall UROP2100, Spring

The report describes the progress of the tasks that has been done in the Spring semester of 2022 and plans to be tried out in the incoming Summer. Integration of the advanced digital information and algorithmic means with modern governance has been a fuzz and one of the tasks of modernization of modern governments in terms of their endeavours endeavors in bolstering governance and state capacity. For mainland China (referred to as China thereafter) as a party state, an increasingly firm and responsive grip of the societal affairs and more importantly, a pre-emptive approach to forestall the protest potentials have been adopted with the assistance of pervasive implementation of digital supervision facilities like closed-circuit television (CCTV) and manual complements entrenched in the grassroot governance system like Community Grid-style Management (CGSM), (shequ) wanggehua guanli. The paper in the first part will recapitulate the related literature regarding the CGSM framed against the backdrop of overall supervision and governance in China, and then the text analysis as a growingly prominent tool in analysing the text data an incorporate it into the regression and big data. With the Quanteda package in R language, the paper shows preliminarily the summary statistics of the text dataset and generate word cloud for visual demonstration to represent the baseline landscape.

213

School of Humanities and Social Science Division of Social Science

Identifying Protest Events in China with Social Media Data Supervisor: ZHANG Han / SOSC Student: LI Xinyue / GCS-SOSC Course: UROP1100, Summer

The use of Internet exploded since 1990s in China with tremendous growth in social media, which built up the channel and platform for Chinese citizens to express their opinions and grievances. Leaders emerge as people with same interest gather through social media and protest for their rights. For fear of large-scale protests, the government is putting more surveillance and control on the social media. Online posts related to sensitive topics are shielded to the public. This paper analyzes social media dataset about protests in China. The statistical analysis reveals that the three major problems underlying in Chinese society are land acquisition, home property rights, and unpaid wages. Violent or disruptive actions are most common forms of protests, as people find no solutions to solve their problems, and companies or governments tend to use extreme method to suppress the protests. The research result about social media and protests is worthy, as it has implication for the social problems rooted in China and the relationship among government, citizens, and social media.

How to Use Machine Learning Predictions in Regression Model Supervisor: ZHANG Han / SOSC Student: CHO Young Beom / SOSC Course: UROP1100, Spring

The paper is the progress report on UROP 1100 project, How to use Machine Learning Predictions in Regression Model supervised by professor Han Zhang. The objective of the project is to implement the CCER package written in the R to Python interface. The paper first shows the literature that was mainly referred to understand the conceptual and mathematical side of the correction model, then briefly explains the Python libraries which are used and their functionalities. The paper reports the current progress, ongoing issues of discrepancies in the optimization function between R and Python, and deciding which Bayesian modeling libraries to use. Finally, the paper concludes by introducing further tasks that can be developed to improve the package, in both the statistical modeling and ML prediction error side.

214

Abstracts of UROP Projects 2021-22 Interdisciplinary Programs Office

Interdisciplinary Programs Office Division of Emerging Interdisciplinary Areas

Division of Emerging Interdisciplinary Areas Algorithms and Games in Android Devices Supervisor: HUI Pan / EMIA Student: CHOI Cheuk Woon / COMP Course: UROP1100, Fall

In this project, I have been tasked to create an app which provides practises on learning Chinese vocabulary. The app itself has several minigames and practises practices inside, which includes writing out the chinese words, finding radicals (部首) of a word, choosing the correct vocabulary for a certain sentence, playing a maze game and so on. The game engine that I used is Unity, a game engine that allows creators to write c# scripts, utilize utilize components and elements like buttons, canvas, etc., and create apps that are accessible by multiple platforms such as windows, android, iOS, etc.

Algorithms and Games in Android Devices Supervisor: HUI Pan / EMIA Student: GAO Xuanrong / DSCT Course: UROP1100, Spring

This report statistically analyzes all Blockchain transactions happening within one day all over the world on Ethereum platform. As all transaction data are stored in a shared and transparency database, users can easily access information about any transaction or block. Therefore, transactions are collected through some Application Programming Interface endpoints, followed by data cleaning and processing to exclude outliers and focus on the majority part. Millions of addresses and transactions form a complex network, while some big companies having a large number of transactions are centers of the graph. Most of the users are individuals with only several transactions. Based on this observation, we develop some statistical analysis employing precise and specific data.

Algorithms and Games in Android Devices Supervisor: HUI Pan / EMIA Student: HUANG Beichen / COMP Course: UROP1100, Spring UROP2100, Summer

We want to do a tutorial for avatar creation that can be useful for prototyping. The platform we use is Unity. We focus on the animation mainly in three parts: Trigger animation with one button, adding sound effects and Navigation Mesh. For now, we have achieved these results: 1. Able to trigger Animation with one button. 2. Able to make each line of avatar stand up in order. 3. Able to play background sound. 4. Able to play corresponding sound effects of each animation. 5. Able to let avatar walk from start point to desired destination. 6. Able to detect the status of avatar(arrived or not arrived) and play different animation.

216

Interdisciplinary Programs Office Division of Emerging Interdisciplinary Areas

Algorithms and Games in Android Devices Supervisor: HUI Pan / EMIA Student: WANG Yixuan / MAEC Course: UROP1100, Spring

This report statistically analyzes all Blockchain transactions happening within one day all over the world on Ethereum platform. As all transaction data are stored in a shared and transparency database, users can easily access information about any transaction or block. Therefore, transactions are collected through some Application Programming Interface endpoints, followed by data cleaning and processing to exclude outliers and focus on the majority part. Millions of addresses and transactions form a complex network, while some big companies having a large number of transactions are centers of the graph. Most of the users are individuals with only several transactions. Based on this observation, we develop some statistical analysis employing precise and specific data.

Algorithms and Games in Android Devices Supervisor: HUI Pan / EMIA Student: XU Xiao / COMP ZAU Ka Ming / CPEG Course: UROP2100, Fall UROP2100, Fall

To begin with, we needed to collect the data before analysing. Initially, we are given 80 accounts id to collect information. The first has been completed so as to guide us. The first row of the table consists of two colours, yellow and green. The yellow section shown in the Table1 includes information from the Instagram profile page, we collected information such as url, description, IG links, etc.

Machine Learning on Wearable Devices Supervisor: HUI Pan / EMIA Student: AGARWAL Sahil / COMP Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

This research study focuses on the various techniques used to build 2 dimensional and 3 dimensional avatar models as part of the HKUST Metaverse on wearable devices that are making use of increasing amounts of complex data. This paper highlights the methodologies I have used to develop a variety of different avatar models by myself and with the help of other UROP students. The primary purpose of the meta verse and avatars is to ensure an online presence of individuals that substitutes their physical need to travel. The meta verse also allows us to connect the virtual world to the physical world. As a result, there are data science and analytical factors that go into preparation of the models. This is where the UROP team including myself have used multiple techniques to analyse and build realistic avatars and generate a website that displays them.

217

Interdisciplinary Programs Office Division of Emerging Interdisciplinary Areas

Machine Learning on Wearable Devices Supervisor: HUI Pan / EMIA Student: LAM Sing Yu / CS Course: UROP1100, Fall

This research study focuses on the various techniques used to detect emotions on wearable devices that are making increasing use of EXtended Reality (XR) technologies such as Augmented and Virtual Reality. The primary purpose of XR technologies is to connect the virtual world to the physical world. Detecting emotions plays a key role in streamlining said connectivity. In this work, we aim to evaluate the state-of-the-art techniques used to detect emotions in XR environments or that can be used in such scenarios.

A De-polarization System for Social Media Supervisor: HUI Pan / EMIA Student: LAU Ching Ming Samuel / COMP Course: UROP1100, Spring

Nowadays, the social media algorithms are designed in a way that enforce personalisation and hence there is a tendency for the echo chambers to be built(Liu et al., 2021). This situation could be manifested in political news, as people only trust and agree with the stance of the media that they tend to watch. To prevent this situation, a tool is proposed to de-polarize news article by recommending news with similar content but different stance and bias. News with various stances will be recommended instead of all opposite stances to attract users to use this tool. In this report, the entire flow of the tool will be discussed, as well as the results and future work.

A De-polarization System for Social Media Supervisor: HUI Pan / EMIA Student: NGAMMANKONGTANGKIT Suttinai / CPEG Course: UROP1100, Spring

With the current Ukraine-Russia war, it affected the social media widely. As we are in a digital world, people tend to receive their information through Internet which raises to the misinformation via social media. This is to discredit the other side and position themselves as the right side. This report aims to present the research on Ukraine-Russia misinformation starting from searching for rumor through official sources, organizing gathered information in the excel sheet, identifying keywords for each rumor, and finally group them into different categories. Finally, the report analyzed the information collected and provided conclusions for this research.

218

Interdisciplinary Programs Office Division of Emerging Interdisciplinary Areas

Can We Build a Cloud from the Crowd ? Cloud Computing on Smartphones Supervisor: HUI Pan / EMIA Student: CHAN Kei Chi / COMP Course: UROP1100, Spring

The proliferation of mobile devices and the advancement of mobile devices’ capabilities has opened new methods for handling computationally intensive tasks. The latest mobile phones are capable of rendering complex graphics in real-time for games with their advanced hardware. What if we can gather these capable tiny computers and have them solve a complex task by collaborating? In places like educational institutions and companies where people are crowded, and thus have many phones available, research that requires taxing computation like machine learning and AR is often performed. This research focuses on unleashing this potential and finding a new way to handle these tasks fast and cheaply.

This report proposes a task scheduler that can coordinate mobile devices to tackle complex tasks together. It captures the structure of the task as a graph, breaks it down, and assigns them to devices over the network. It also automatically prioritizes sub-tasks that could become a bottleneck and ensure they execute at the fastest speed. To be tolerant of network errors, it also has a recovery mechanism that can restart any subtask that did not return successfully.

The main goal of this scheduler is to finish the task at the fastest speed by distributing sub-tasks to a scalable network of mobile devices. And this report will delve into its design choices, the development progress, and future steps.

Augmented Reality on Wearable Devices Supervisor: HUI Pan / EMIA Student: CHE Siu Hei / COMP Course: UROP1100, Spring

This report demonstrates a method for hand gesture recognition using machine learning models. Instead of using camera tracking which is what most commercial virtual reality or augmented reality headsets use for hand gesture recognition, a glove-based hand gesture recognition is not affected by the limitation of a camera, as it can still work when user’s hands are occluded or outside of the camera’s field of vision. Using an IMU glove for data collection, the data is then preprocessed and fed into two kinds of machine learning model, CNN and FNN. After that, a real-time test is conducted to assess the performance of the models. The results indicated that a FNN model is better at recognizing gesture than a CNN model.

219

Interdisciplinary Programs Office Division of Emerging Interdisciplinary Areas

Human Faces Generation Using GAN Supervisor: HUI Pan / EMIA Student: TONG Tsun Man / SENG Course: UROP1100, Summer

The project “Human Faces Generation using GAN” is conducted during the Summer Semester in 21-22 term under the supervision from Professor Hui PAN, and PhD student Steve Weipeng ZHUO. Several body generation projects were investigated and put into practice for generating a three-dimensional human object with only the frontal image, and attempts to implement with multiple images input, such as PIFu, HeadNeRF, Neural Head Avatars and Avatar SDK that make use of methods such as Deep Learning and Generative Adversarial Network (GAN). The project was very rewarding in terms of having a deeper understanding of Computer Vision, Machine Learning and Deep Learning.

This report mainly focuses on these projects’ implementation, difficulties, their final results, their advantages and disadvantages.

Human Faces Generation Using GAN Supervisor: HUI Pan / EMIA Student: ZHANG Chenyu / SENG Course: UROP1000, Summer

In the UROP 1000 project, I worked on the project: Avatar Creation for the Meta Opening for HKUST Guangzhou Campus in September, with Professor HUI Pan as my supervisor. I mainly went through GitHub programs, including PIFu, RigMe, PIFuHD and Avatar SDK, that are used for the purpose of generating 3D model out of one front photo of a person. I modified the python environment, executed the programs, and finally compared the results generated from the programs. As a result, I found the Avatar SDK is the most satisfying, although there are still a few parts unsatisfying. I also helped Steve modify the website for receiving photos of presidents and mark the received photos.

220

Interdisciplinary Programs Office Division of Environment and Sustainability

Division of Environment and Sustainability Game-theoretic Analysis of Markets Supervisor: CHAN Jin Deng Keith / ENVR Student: CHEUNG Chun Bon / EVMT Course: UROP1000, Summer

Ultimately countries want to develop policy frameworks that maximize the ‘impact’ of the financial sector on the fight against climate change (ceteris paribus on financial returns). Governments are making efforts to create policy frameworks to help mobilize finance for climate action. Standardized eligibility criteria for labelling financial instruments as ‘green’ are widely considered as critical for providing credibility and confidence to scale this burgeoning market. Countries with strong enforcement capacity can accomplish this by using a TSC approach, while countries with weaker enforcement oversight are better served by using a white list (WL) approach. This paper show that as countries with weaker enforcement improve their external review capacity, they can shift from a WL to Technical Screening (TSC) approach.

Game-theoretic Analysis of Markets Supervisor: CHAN Jin Deng Keith / ENVR Student: LEUNG Man Ting / MAEC Course: UROP1000, Summer

In the green bond market, issuers can seek external reviews to validate their green claims. This paper studies the effectiveness of different review types in information provision by comparing their unique characteristics. A review of different categorizations of “greenness” by reviewers is also performed. This could act as an independent variable in a model with ‘effective information provision’ as the dependent variable. Our finding shows that second-party opinions (SPO) and green bond scoring particularly excel in this area relative to other review types. Future analysis would have to identify what constitutes ‘effective information provision’ for investors, and thus identify the remaining independent variables of the model.

Interrogating Fintech-Enabled Solar Power as 'Energy Democracy' in Hong Kong Supervisor: DELINA Laurence Laurencio / ENVR Student: LAM Rainbow Yi Hung / EVMT WONG Ka Ying / EVMT Course: UROP4100, Summer UROP1000, Summer

In the Summer of 2022, this UROP continued to examine the statuses and potentials of a fintech-enabled renewable energy market and its relationship with energy democracy in Hong Kong. The students completed a journal article template for a Science Talks submission. Science Talks is an open access video journal that publishes original, peer-reviewed scientific talks from experts in the field. The video journal was published in October 2022 as Delina, L. L., Lam, R. Y. H., Tang, W. S., & Wong, K. Y. (2022). Hong Kong's “future” energy transition-digitalization-financialization triangle: Actors and impacts. Science Talks, 4.

221

Interdisciplinary Programs Office Division of Environment and Sustainability

Interrogating Fintech-Enabled Solar Power as 'Energy Democracy' in Hong Kong Supervisor: DELINA Laurence Laurencio / ENVR Student: TANG Wing Shun / EVMT WONG Ka Ying / EVMT Course: UROP1100, Fall UROP2100, Fall

In the Fall of 2021, this UROP continued to examine the statuses and potentials of a fintech-enabled renewable energy market and its relationship with energy democracy in Hong Kong. The students re validated the primary data they collected from key informant interviews and coding this qualitative data using MAXQDA. MAXQDA is a software package for qualitative methods. The outputs for this UROP included a coded dataset for analysis and publication in a journal.

Japan-funded Energy Projects in Southeast Asia Supervisor: DELINA Laurence Laurencio / ENVR Student: WANG Zhiping / SSCI Course: UROP1000, Summer

This UROP looked at Japan’s investments in Southeast Asia’s energy sector development. In Summer 2022, the student conducted a literature review. Some 149 articles were read and analyzed, suggesting the many complex reasons for Japan’s investments in the energy sector.

Japan-funded Energy Projects in Southeast Asia Supervisor: DELINA Laurence Laurencio / ENVR Student: YEUNG Sen / SSCI Course: UROP1000, Summer

Japan has a long-standing relationship with countries in Southeast Asia, financing and aiding them in building infrastructures. This UROP project assessed Japan’s energy investments in this world region from 2000 to 2020, paying closer attention to reviewing the literature. The review focused on how Japan invest in Southeast Asia's energy sector, and why it invests in this world region.

222

Interdisciplinary Programs Office Division of Environment and Sustainability

Air Quality Measurement Techniques Supervisor: GU Dasa / ENVR Student: CHAN Ching Ying / BIBU Course: UROP1000, Summer

The team has measured the level of ambient volatile organic compounds (VOCs) in the HKUST supersite. There are also a lot of collaboration work with the other teams of ENVR and the EPD, focusing on wind, humidity and so on. The team collects samples in the supersite 3 times a day for doing analysis. Air samples will be collected in a canister which is convenient for us to further speculate the compounds in the ambient air. Air pollution is a serious problem in Hong Kong. It brings physical threats to humans, especially respiratory diseases. Various kinds of air pollutants have been created and dispersed in the air. There are several groups of VOCs commonly found in the air samples. They are alkanes, cycloalkanes, alkenes, aromatics, monoterpenes, alkyne, halogenated, alkyl nitrates and sulphur compounds. In this report, we will focus on alkanes, alkenes and alkynes.

Machine Learning for Environmental Applications Supervisor: LAU Alexis Kai Hon / ENVR Co-supervisor: FUNG Jimmy Chi Hung Student: LEI Qianqian / MATH-SFM Course: UROP1100, Fall

In order to reduce the losses caused by climate impacts, people need to be equipped with basic knowledges of possible climate risk. A well designed database, from which people can access all the risks and be prepared for the climate risk, will be helpful. In this progress report, it first introduce the whole idea of the project, why and what we are going to do. In addition, is our basic framework of the project. Through this semester, what I have done is literature review. I will further give a thorough introduction of the literature review, which will be foundation for our future process.

Machine Learning for Environmental Applications Supervisor: LAU Alexis Kai Hon / ENVR Co-supervisor: FUNG Jimmy Chi Hung Student: WONG Chi Ho / DSCT Course: UROP1100, Spring

IoT sensors are prominent nowadays for detecting air quality, which let the public reference it or helps the commercial buildings to operate and do maintenance efficiently. However, anomalies may happen in those sensors and cause confusion before human intervention. We may meet the dilemma that whether should professionals trust the extraordinary data and respond with some follow-up actions on the system. Therefore, it is crucial to accurately detect anomalies with the help of a deep learning model MAD-GAN, which can model the temporal and spatial contextual information for better detection. We achieved around 80-90% accuracy in determining anomalies on the HKUST IoT Sensor dataset.

223

Interdisciplinary Programs Office Division of Environment and Sustainability

Visualization of Tree Failure in Super Typhoon Mangkhut Supervisor: LU Zhongming / ENVR Student: RATTANAMANEEJARAT Thananun / EVMT Course: UROP1100, Fall

After the damage done from Hong Kong’s most intense Typhoon in its history, tens of thousands of trees experienced tree failure, falling and causing damage to properties and people. Therefore, the visualization of tree failure data using Geographical Information Systems (GIS) such as ArcGIS can be a tool for research into the various reasons and factors that can contribute to making tree failure more or less likely.

A tangential exploration of the various ways that ArcGIS can be used to portray data is also shown, along with some literature research into Sky View Factor (SVF), which is a property of urban landscapes that can affect wind speeds and climatological conditions at a local level.

Visualization of Tree Failure in Super Typhoon Mangkhut Supervisor: LU Zhongming / ENVR Student: WONG Mei Yu / EVMT Course: UROP1100, Fall

The project tries to measure the tree failure risk in Hong Kong built-up area using a sample of tree failure during super typhoon Mangkhut. ArcGIS is selected as the geographical information system software used for modelling to visualize the tree fall situation because of Mangkhut. After considering the related factors on tree falling, the impact of the geology of Hong Kong is chosen as the focus of this study. Through analysis of the relationship between tree falling and geographical features, it is hoped to provide insight into the management of urban trees and to increase the resilience of Hong Kong under extreme weather events.

Affordable, Sustainable Fashion!? Supervisor: SAUERWEIN Meike / ENVR Co-supervisor: MATUS Kira / ENVR SINGH Renu / PPOL Student: SUTEDJA Amrita Saraswati / ACCT Course: UROP2100, Fall UROP3100, Spring

This study aims to analyze pandemic effects on fashion consumption attitude and behavior as part of a ninecountry/ region study, and the “Affordable, Sustainable Fashion!?” UROP project is in charge of the Hong Kong survey distribution and analysis. Based on this survey, we found several trends and areas for further investigation which prompted us to conduct a second survey, broadening the scope to the whole Hong Kong society and collaborating with Redress, an environmental NGO in Hong Kong, to quantify fashion waste flows in Hong Kong. We aim to look into the sustainable fashion intention-behavior gap as well as end-of-life options of fashion items to get a better understanding of the sustainable fashion scene in Hong Kong. Both the second survey and Redress collaboration are currently underway, and we are confident that upon completion, the results can enrich the literature around the sustainable fashion field, especially in Hong Kong and Asia.

224

Interdisciplinary Programs Office Division of Environment and Sustainability

Affordable, Sustainable Fashion!? Supervisor: SAUERWEIN Meike / ENVR Co-supervisor: MATUS Kira / ENVR SINGH Renu / PPOL Student: HALIM Vanessa Theodora / CPEG MAK Tin Nok / IPO PANJETA Simarpreet Kaur / MARK Course: UROP1000, Summer UROP1100, Summer UROP1000, Summer

The “Affordable, Sustainable Fashion!?” UROP project aims to investigate the general public’s fashion consumption pattern under the influence of the COVID-19 Pandemic in Hong Kong, as well as the government actions and/or private initiatives towards the textiles waste problem mainly across five countries/regions nearby Hong Kong. Literature reviews on government data and local community initiatives are conducted to understand more about the amount of textile waste, recycling rate and end-of-life options of clothing items in different Asian societies. Additionally, as a continuation of previous findings to understand more on consumers’ fashion consumption pattern since the pandemic, a second survey with Chinese translation had been developed and launched to broaden the previous scope of Hong Kong University Campus to the entire Hong Kong society. The survey result, which will be analyzed in the future, can provide useful insights regarding sustainable fashion consumption in Hong Kong.

Apps and Tools to Encourage Responsible, Sustainable Consumption Supervisor: SAUERWEIN Meike / ENVR Student: YEUNG Man Lai / CHEM Course: UROP1000, Summer

Exploring new methods that can instil the mindset towards sustainability consumption (SC) and cultivate a habit of mindful consumption to the public is necessary for the near future since issues of resource limitation are intensifying. While there are more and more educational classed and initiatives for students to learn about SC in schools or university there are limited ways to reach general citizens to acquire knowledge or change their attitude to induce behaviour change. Earlier studies have investigated how playing games can increase awareness and knowledge of students on particular issues, such as sustainability, yet there is little evidence that such games could be accessed by the general public and have a similar educational impact if played outside a class setting. Furthermore, there is a lack of statistical evidence for such studies. Hence, this study focuses on how effectively serious gaming can increase the awareness of sustainable and collaborate consumption (CC), comparing game-play in an educational context with a casual game play setting. The findings show that serious gaming can raise the awareness of CC and SC in both casual play and serious gaming, while educational debriefing after gaming could maximize the effectiveness of raising SC awareness among participants.

225

Interdisciplinary Programs Office Division of Environment and Sustainability

Driving Behavioural Change of BYOBottle by Smart Data and Innovation (SSC Smart Meters Project) Supervisor: SAUERWEIN Meike / ENVR Student: FAN Xingyi / EVMT OUDOMYING Por / EVMT Course: UROP1100, Fall UROP2100, Spring

Last semester, we installed interventions on some water dispensers around campus to test the significance of hygiene and safety perception on water consumption. During this semester we analyzed the data collected from water dispensers and quantified the effectiveness of our intervention by observing the refill trends throughout the semester, and comparing the percentage change of refills for test and control sites. In addition, we have launched a post-intervention survey and conducted qualitative and quantitative analysis on the responses collected, to investigate how the intervention has changed students' perceptions towards the water dispensers on campus. In the end, recommendations are made for future improvements to encourage more water dispenser usage and water consumption on campus.

Decarbonization 2050 with Green Energy - An Educational Game Supervisor: SAUERWEIN Meike / ENVR Student: LAU Chi Choi / EVMT LOK Chuen Leung Ferdinand / EVMT Course: UROP1100, Fall

With the rising importance of carbon-neutrality by 2050, we would like to promote this concept in a simpler and easier way to the public in a bid to enhance their environmental awareness. Therefore, the card game project, Decarbonization 2050, is carried out.

Even though we have conducted various market research to design the prototype from a more accurate angle of players, we realized that the first draft of our prototype was not player-comprehensible enough. The lack of interaction and variation bored our demo players. Noted the above concern, we are bettering the set up and rules of Decarbonization 2050. It is believed we would be able to improve the game so that it could be both educational and attractive.

226

Interdisciplinary Programs Office Division of Environment and Sustainability

Extreme Weather Under Global Warming Supervisor: SHI Xiaoming / ENVR Student: TAN Hanzhi / SSCI Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

Continuing from the last UROP Project, further research concerning to the accuracy of different models simulating Mei-yu pattern have been proceeded. To find its relationship with the modelling of Westerly wind around Tibetan Plateau, climate models from CMIP6 has been more detailed analyzed in particular periods of an annual climate phenomenon. Thus, the methods could be found more specifically. Here, Mei-yu can be understood as a macro concept of typical precipitation responds under the East Asian Monsoon, which can be further divided by “pre-Mei-yu” (Apr-Jun) and “main-Mei-yu” (Jun-Aug). As the latter one is likely to be disturbed by some small-scale weather phenomena bursts in midsummer, there should be a method to eliminate such noises.

227

Interdisciplinary Programs Office Division of Public Policy

Division of Public Policy Facilitating Data-Driven Innovation for Sustainability: Opportunities and Challenges in Public Policy Supervisor: YARIME Masaru / PPOL Student: SHREEHARSH Aastha / MAEC Course: UROP1100, Fall UROP2100, Spring

With the ever-increasing complexity of the problems faced by the global community, it has become vital to explore the significant potential of data-driven innovation in solving the same. This project seeks to assess one such innovation—the use of blockchain-based digital certification and verification of academic credentials for higher education. This progress report follows from a previous discussion of existing case studies, the obstacles in their implementation and their potential benefits. It further explores the opportunities and challenges that a wide-scale implementation may present from the public policy perspective.

228

Abstracts of UROP Projects 2021-22 Institutes and Centers Institutes and Centers Institute for Emerging Market Studies

Institute for Emerging Market Studies Trade and Investment under China's Belt and Road Initiative - Implications for Hong Kong Supervisor: TRITTO Angela / IEMS Co-supervisor: PARK Albert Francis / IEMS Student: JIANG Yuxuan / ECOF Course: UROP2100, Fall

The recent years have witnessed the growth of Chinese Development Assistance, especially under the COVID-19 pandemic. Therefore, it is worthy to investigate the further impacts of China’s aid on its international status worldwide. The project examined the voting decisions of resolutions in the United Nations Human Rights Council (UNHRC). At the current stage, the researcher conducted data analysis to explore China’s voting preferences in controversial resolutions, and summarized several determinants of Chinese Foreign Direct Investment, foreign aid, medical aid, and UNHRC voting decisions to find potential factors to be included in the project’s models. The tests results showed that significant differences existed in China’s voting decisions towards controversial resolutions and normal resolutions. Furthermore, eight dimensions of factors are potentially helpful in model construction, which are drafted as part of the literature review.

Trade and Investment under China's Belt and Road Initiative - Implications for Hong Kong Supervisor: TRITTO Angela / IEMS Co-supervisor: PARK Albert Francis / IEMS Student: WU Hongsen / GBUS Course: UROP3100, Fall

Providing countries in need with medical assistance under the COVID-19 pandemic, including medical goods donation, vaccine donation and vaccine sales, has become an important component of China’s foreign policy. This research project aims to understand the determinants of the distribution of Chinese medical assistance, especially how political alignment between China and the recipient countries affect the distribution. This progress reports highlights how we constructed the quantitative models, and how we collected data for different variables.

230 香港科技大學 THE HONG KONG UNIVERISTY OF SCIENCE AND TECHNOLOGY