Undergraduate Research Opportunities Program (UROP) Proceedings 2019-20

Abstract

香港科技大學 The Hong Kong University of Science and Technology Undergraduate Research Opportunities Program PROCEEDINGS 2019-20

Message from the President i Sharing by Students ii UROP Overview 2019-20 iii Mr. Armin & Mrs. Lillian Kitchell UG Research Awardees 2020 iv UROP Sponsorship Recipients 2019-20 v-vi Abstracts of UROP Projects – 2019-20* School of Science (CHEM, LIFS, MATH, OCES, PHYS) 2-50 School of Engineering (CBE, CIVL, CSE, ECE, IEDA, ISD, MAE) 52-112 School of Business & Management (ACCT, DBM, ECON, FINA, ISOM, MARK, MGMT) 114-144 School of Humanities & Social Science (HUMA, SOSC) 146-162 Interdisciplinary Programs Office (ENVR) 164-170 Institutes & Centers (IEMS) 172-173

* Abstracts from each School are listed first by alphabetical order of Department code, and then by alphabetical order of 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 UROP1100, with approval by project advisors)

UROP 3100 Undergraduate Research Opportunities Program Series 3 (1 credit, offered throughout the year; prerequisite is pass in UROP2100, with approval by project advisors)

UROP 4100 Undergraduate Research Opportunities Program Series 4 (1 credit, offered throughout the year; prerequisite is pass in UROP3100, with approval by project advisors) Message from the President The 2019-20 academic year is a year full of unprecedented events and challenges. Despite the social unrest and the outbreak of the COVID-19 pandemic, the whole HKUST community has been working hard to keep the learning activities for our students on track. This volume of proceedings showcases the collaborations and achievements of our faculty supervisors and students in the Undergraduate Research Opportunities Program (UROP) in 2019-20.

The UROP continued to play a valuable role for our undergraduate students to have an early exposure to research and to explore their interests across different academic disciplines. I am glad to see another year of success for the UROP, with continuous support from our faculty members offering a great number of interesting and diverse projects. The UROP continues to attract hundreds of undergraduate students to take their first steps in their research journeys. I believe that students would find their research work challenging yet inspiring, stumbling at times yet always rewarding.

With every challenge we face and eventually overcome, there is always an opportunity to open and a new lesson to learn. As a young university, the HKUST has always been striving to advance learning and knowledge through teaching and research. By offering the UROP and many other learning opportunities, we hope to nurture a diverse and conducive culture to help our students develop their talents and achieve their aspirations.

Going forward, we will continue to bring together faculty members and undergraduate students to break new grounds in research, to take new steps that no one has ever dreamt of, and to go beyond academic boundaries for societal impact. We hope you will join us in this rewarding journey.

Prof. Wei Shyy President HKUST

I

Sharing by Students

SHIHUA Mingzhi BEng in Aerospace Engineering

Life is all about making choices, and enrolling UROP is definitely one of the most important and wise choices I have made in my life (choosing HKUST must be another such perfect choice). I started UROP with no prior knowledge in this field, as a field of research is often very narrow and in-depth which is totally different from classroom education. Fortunately, both my supervisor Prof. Rhea and other graduate students helped me from briefing me the essential knowledge and stills, getting me involved in the research itself (instead of repeating other people’s work), performing my research independently, to eventually publishing a high-quality top conference paper. This experience is wonderful because it not only helped me learn state-of-the-art knowledge in this professional field, but also gave me a taste of what a scientific research is like. The program shaped me with a sharper mind, the ability of critical reasoning and a better understanding of my future choices to make.

WONG Ting Hei Matthew BSc in Chemistry

The UROP project was an amazing experience. I’ve had a great time at the lab talking to different researchers and learning from them various ways of solving problems. After working for a while, I’ve come to realize that the key to success in research is dedication. Even with the greatest talent in the world, one cannot translate that gift into results without investing a lot of time in the project. Never falter in the face of hardship; instead, persevere and you will see your hard work blossom into a reward far beyond your imagination.

LEE Hojun BSc in Quantitative Social Analysis

UROP summarizes my undergraduate experience at HKUST. I started UROP in my first undergraduate year and continued it almost every semester until graduation. It was first started to acquire extra credits, but it later became my primary impetus to learn and pursue deeper knowledge. I had to learn new programming languages to perform the analysis I wanted to, and I had to take political science and history courses to supplement my understanding of the research context. I think this is what makes UROP so special. UROP is not about earning scores but about proactively learning to become a valuable member of a research team.

ii UROP Overview 2019–20

In the 2019-20 academic year, the Undergraduate Research Opportunities Program (UROP) continued to serve as a unique learning platform for undergraduate students to explore their interested research areas across different disciplines. We are glad to see that participation in the UROP from both faculty members and undergraduate students recorded continuous increase compared to the previous year. Over 200 faculty members offered nearly 500 projects throughout the year, and more than 530 undergraduate students joined the program under the supervision of the faculty advisors.

Through the UROP Support Grant, UROP students can try how to apply for research funding by submitting a joint application with their supervisors. This funding scheme also aims to provide additional financial support to help enhance the continuous development of an existing UROP project, and to encourage a sustained UROP experience for undergraduate students. In 2019−20, 46 valid applications were received and reviewed by the UROP Office and the UROP Advisory Board. In total, 34 projects were awarded with an aggregate funding amount of over HK$536,000.

The 2020 Mr. Armin and Mrs. Lilian Kitchell Undergraduate Research Award and the UROP Faculty Research Award

In 2020, we received 22 nominations from the UROP faculty supervisors for the Mr. Armin and Mrs. Lilian Kitchell Undergraduate Research Award to recognize their students’ outstanding performance in the UROP projects. 10 students were shortlisted as the finalists. After presenting their projects to the UROP Advisory Board, six of them were selected as awardees (one Champion, two First Runner-ups and three Second Runner-ups). The student awardees’ faculty supervisors received the UROP Faculty Research Award as an appreciation for their dedicated support to the program. Due to the outbreak of the COVID-19 pandemic, the UROP Award Presentation Ceremony this year was held online via Zoom on June 5, 2020 and the UROP Week was suspended in 2020. It was our honor to have around 170 guests joining the ceremony. This year, in addition to the project presentations by the top three awardees, student awardees and their faculty advisors were invited to give a speech at the Ceremony.

iii

Mr. Armin and Mrs. Lillian Kitchell Undergraduate Research Award 2020 List of Awardees

Champion SHIHUA Mingzhi Major/Year: AE / Year 3 Supervised by: Professor LIEM Rhea Patricia / MAE Project Title: Designing Novel Efficient and Quiet Amphibious Aircraft for General Aviation

First Runner-up SIM Yi Hang Major/Year: EEGBM / Year 4 Supervised by: Professor CHAU Kevin / ECE Project Title: Projects in Audio Signal Processing – Advanced Automobile Crash Detection by Acoustic Methods

First Runner-up WONG Ting Hei Matthew Major/Year: CHEM / Year 3 Supervised by: Professor SUN Jianwei / CHEM Project Title: Organocatalytic Synthesis of Useful Chiral Molecules

Second Runner-up LEE Hojun Major/Year: QSA / Year 3 Supervised by: Professor KELLER Franziska / SOSC Project Title: Political Astroturfing in South Korea: Examining How the Secret Service Manipulated Twitter during the 2012 Presidential Election

Second Runner-up LEUNG Ziuwin Major/Year: BIOT / Year 4 Supervised by: Professor WU Angela / LIFS Project Title: Interdisciplinary Bioengineering: Microfluidic Device for Cell Manipulation / Culture and Molecular Biology

Second Runner-up & Best Poster Award YU Tsz Tat Major/Year: MECH & SUSE / Year 3 Supervised by: Professor CIUCCI Francesco / MAE & CBE Project Title: Polymer Composite Batteries

Iv

UROP Sponsorship Recipients 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.

Students who have been awarded UROP sponsorships in the 2019-20 academic year are listed as follows:

UROP Research Travel Sponsorship (by alphabetical order of student’s surname) Student: LAU Hoi Man Major: PHYS UROP Supervisor: Professor PARK Hyo Keun / LIFS & PHYS Conference: The 64th Annual Meeting of the Biophysical Society Venue: San Diego, California, USA Duration: February 15-19, 2020 Summary of Research Activity: It had been my pleasure to attend the 64th Annual Meeting of the Biophysical Society this February and to present my UROP project “Real Time Nano-meter Accuracy Tracking of Lipid Droplets in Living Cells” in the poster session. The feedback was thrillingly positive and ideas and suggestions from experts in different fields such as molecular modelling, fluorescence biomarker and protein synthesis, were constructive and inspiring. More importantly, the session provided an opportunity to reach for collaboration and confirmed the scientific significance of this project. Besides poster presentation, I also attended various workshops and symposia. Among those edge-cutting researches introduced, I found the single molecule protein sequencing and acoustiofluidic interferometric techniques exclusively impressive.

v

Student: WAN Yuxuan Major: CPEG & MATH-AM UROP Supervisor: Professor CHAU Kevin / ECE Conference: The 147th Convention of the Audio Engineering Society Venue: Jacob K. Javits Convention Center, New York City, USA Duration: October 16- 19, 2019 Summary of Research Activity: In October 2019, I attended the AES 147th Convention held in New York with the help of UROP travel sponsorship. I participated in the UROP for three semesters, spring 2019, summer 2019, and fall 2019, under the supervision of Professor Kevin Chau. Based on the results in summer 2019, I published a paper on the AES Convention and was invited to give an oral presentation. The work proposed an algorithm for onset detection in audio signals based on wavelet transform. In addition to the precious experience for presentation, attending the conference gave me an opportunity to share the research results and interact with other young researchers. Finally, I would like to express my sincere appreciation for the support from Professor Kevin Chau and the UROP Office.

Student: SIM Yi Hang Major: EEGBM UROP Supervisor: Professor CHAU Kevin / ECE Conference: The 178th Meeting of the Acoustical Society of America Venue: San Diego, California, USA Duration: December 2 - 6, 2019 Summary of Research Activity: Modern automobiles are equipped with a variety of safety features, but have yet to widely utilize acoustics and sound, although these can provide valuable information and advance warning of a potential car crash. Research was conducted on a method to utilize acoustic waves from a car or surrounding cars, to predict in real time a car crash in advance. The algorithm uses the discrete wavelet transform, which provides multiple time and frequency resolution and a different perspective on sound. By weighting different components of the sound, events such as honking, tire skidding, and collision can be isolated and detected. In particular, tire skidding can be used to predict a car crash and provide information to a car, which can then activate other safety features such as hazard lights or seatbelt tensioners. Overall, the algorithm performs well and has advantages in sudden onset detection, temporal localization accuracy, and computational cost.

vi

Abstracts of UROP Projects 2019-20 SCHOOL OF SCIENCE

School of Science Department of Chemistry Department of Chemistry

Synthesis of Designed Organic Molecules with Potential Bioactivity Supervisor: DAI Wei Min / CHEM Student: WONG Kim Lok Kimson / CHEM Course: UROP2100, Fall

In this project, progress toward synthesis of the C3–C11 THF fragment of iriomoteolide-13a, a cytotoxic natural product isolated from a benthic dinoflagellate Amphidinium sp., has been made by using (E)-hex-3-enedioic acid as the starting material. The synthesis begun with esterification of the diacid to form the corresponding diester which was reduced by LiAlH4 to afford the diol. The latter was subjected to chemoselective mono-alkylation with 4-methoxybenzyl chloride (PMBCl) to furnish (E)-6-[(4′-methoxybenzyl)oxy]hex-3-en-1-ol along with the bis-alkylation byproduct. Proton nuclear magnetic resonance (1H NMR) spectroscopy and infrared (IR) spectroscopy were used to verify the identity of the synthesized products.

Synthesis of Designed Organic Molecules with Potential Bioactivity Supervisor: DAI Wei Min / CHEM Student: YU Bokshchkhen / CHEM Course: UROP1100, Fall

Sulfa drugs are a group of synthetic antibiotics that contain the sulphonamide functional group. The antibacterial properties of sulfa drugs were first observed in 1932 and they were the first chemical substances to be systematically used to treat and prevent bacterial infections in humans. Although sulfa drugs have mostly been replaced by more effective and less toxic drugs, they still hold great historical significance and were a valuable foundation for the development of superior antibiotics. This project deals with the reactions of sulfonyl chlorides and amides and focuses on the synthesis of the representative member of sulfa drugs, sulfanilamide, as well as, one of its derivatives, sulfassucinamide, using lab techniques and structural analysis tools such as refluxing, vacuum filtration, steam distillation, TLC, and NMR spectroscopy.

Synthesis of Functionalized Organic Molecules Supervisor: DAI Wei Min / CHEM Student: KWOK Chui Ting / CHEM Course: UROP1100, Fall

Chlorpromazine is the earliest antipsychotic drug. It has built the proper scientific knowledge that mental illnesses are only a kind of diseases and can be cured by drugs. Many antipsychotic drugs have been developed based on chlorpromazine to meet tremendous demand of the society. This project aims to synthesize chlorpromazine by comparing the methods reported in the literature and learn how to analyze the reaction mixture. Starting from 2-chlorophenothiazine and N,N-dimethyl-3-chloropropylamine, the reaction has been examined under the conditions reported by Wang and Tang in a 2012 patent and by Gallon and co-workers in a 1985 article, respectively. With the reference to an authentic sample of chlorpromazine, the reaction product was confirmed.

2

School of Science Department of Chemistry

Synthesis of Functionalized Organic Molecules Supervisor: DAI Wei Min / CHEM Student: MA Wing Yin Timothy / CHEM Course: UROP1100, Fall

Aspirin (acetylsalicylic acid) is one of the most prevalent and best known over the counter drugs around the world. It is used as a relief of pain and fever, and to reduce inflammation. This report focuses on the synthesis of aspirin from a naturally occurring compound – oil of wintergreen (methyl salicylate), by a two-step reaction that involves a hydrolysis of the methyl ester group on the oil of wintergreen to obtain salicylic acid, and then an acetylation on the phenolic hydroxy group of salicylic acid using acetic anhydride. The structures of the reaction intermediates and product were confirmed using proton nuclear magnetic resonance (1H NMR) and melting point determination was used to check the purity of the synthesized compounds.

Construction and Application of Surface Enhanced Raman Spectrometer in Biomolecules Characterization Supervisor: HUANG Jinqing / CHEM Student: LIU Sum Yee Roxanne / CHEM Course: UROP1100, Fall In this paper, my teammate and I perform Raman spectroscopy on a piece of silicon sample. At the beginning, we design and construct a mini Raman spectrometer with 780 nm laser source. We then test its effectiveness by measuring the Raman signals of silicon sample and revise and recalibrate the design. Although the emission spectrum does not exhibit a satisfactory result, we achieve to detect Raman signal by the end of the UROP project.

Construction and Application of Surface Enhanced Raman Spectrometer in Biomolecules Characterization Supervisor: HUANG Jinqing / CHEM Student: YAN Qiaolin / CHEM Course: UROP1100, Fall

Raman spectroscopy is a crucial technique for molecule characterization in chemistry labs. The method targeting spectrum analysis of sample surface is increasingly gaining popularity as it is quick, robust and non-destructive. The basic principles of Raman spectroscopy are first reviewed and two basic setups using crystalline silicon as sample to explore the basic theory of Raman spectroscopy are then attempted. The obtained Raman spectrum of sample using set up 1 is too weak to be conspicuous and that of setup 2 presents significant discrepancy from expected results, which requires further investigation. Possible explanations are proposed and potential aspects of improvement in setup design are discussed.

3

School of Science Department of Chemistry

Single Molecule Manipulation and Characterization of Biomolecules Supervisor: HUANG Jinqing / CHEM Student: CHAU Tin Yuet / CHEM Course: UROP1100, Spring UROP2100, Summer

BSA and lysozyme are globular proteins, their secondary structures is well known from X-ray crystallographic studies, while alpha-synuclein is a natively unfolded proteins which do not adopt a unique conformation. Although it is a disordered protein, studies show that different environmental changes induce alpha-synuclein to adopt predominantly alpha-helical or beta-sheet conformations. It is found that the secondary structure in alpha-synuclein becomes predominantly beta-sheet in 25% methanol and largely alpha-helical in 1% SDS, there is no noticeable predominant secondary structure shown in water. Thus, the investigation on the secondary structures of the globular proteins through Raman spectroscopy and the connection between their structures and functions potentially offers insight on the amino acid sequence and functions of alpha synuclein.

Single Molecule Manipulation and Characterization of Biomolecules Supervisor: HUANG Jinqing / CHEM Student: LEE Yung Chi / CHEM Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer

Alpha-synuclein is a protein that is believed to be linked with Parkinson’s disease. Raman spectroscopy and singlemolecule force spectroscopy are powerful techniques that can be used to study the structure of proteins in different environment, and thus be useful for probing the structure and folding mechanisms of alpha-synuclein in different physiological conditions and give insights in Parkinson’s disease research. In this stage of study, we recorded normal Raman and Surface-Enhanced Raman Spectroscopy (SERS) spectrum of standard proteins BSA and lysozyme for comparison with previously recorded alpha-synuclein spectrum in different environments. Single-molecule force spectrum of alpha-synuclein is also recorded and evaluated.

Medicinal Chemistry on Novel Type I1/2 ALK Inhibitors for Combating Drug-resistant Mutants Supervisor: HUANG Yong / CHEM Student: WONG Tsz Wang / CHEM Course: UROP1100, Summer

The synthesis of medically useful compounds is usually done by total synthesis. This results in many steps and gives a low yield for desired product. It is proposed that the borylation selectivity of aromatic esters can be achieved by using ligands with a specific length, shape to direct the metal catalyst to the desired insertion site. In this project, the synthesis of 2-Methylbenzoxazole-5- boronic Acid Pinacol Ester, one of the key materials for the synthesis of ALK Inhibitors, from 2- Methylbenzoxazole using metal-catalyzed selective CH remote borylation of Benzoxazoles is studied. Totally 4 ligands were designed and 2 of them have been synthesized using the optimized reaction condition. Nuclear Magnetic Resonance (NMR) Spectroscopy, Gas Chromatography/ Mass Spectrometer (GC/MS) are using to confirm the quality, quantity, and purity of the synthesized compounds. 4

School of Science Department of Chemistry Application of Molecular Orbital Theory to Transition-metal Complexes Supervisor: LIN Zhenyang / CHEM Student: CHAN Hok Tsun / SSCI Course: UROP1100, Summer

Density functional theory (DFT) calculations were carried out to study the reaction mechanisms of intermolecular displacement on cyclopropyl carbinol derivatives to give acyclic molecules. Our investigation focuses on the pathways for the transformation of cyclopropyl carbinol derivatives to acyclic products and examines the origin of regioselectivity involved in the transformation. Through our calculations, we found that intermolecular nucleophilic substitution occurs via a nucleophile-for-OH- SN2 mechanism at the quaternary carbon stereocenter. The stability of the transition states is responsible for the regioselectivity.

Deep Learning in Synthesis Planning Supervisor: SU Haibin / CHEM Student: DO Van Quyet / DSCT ZOU Zhiyan / SSCI Course: UROP1000, Summer UROP1000, Summer

Synthesis planning has become very important in many fields of chemistry, including drug discovery. The involvement of machine learning models in retrosynthesis systems raises the need in more comprehensive reaction information to have better support. That can be achieved by reaction classification. Here we present a method to classify reactions. We step-by-step extracted reaction core (i.e the mechanism of reaction) of chemical reactions from a data set derived from U.S. patents, then base on that to group those reactions. Our work results in a good classification, confirmed senior chemists. Besides, a new type of fingerprint has been constructed for similarity calculation and visualizing reactions in the database with the expectation of discovering some hidden rules.

Deep Learning in Synthesis Planning Supervisor: SU Haibin / CHEM Student: MA Wing Yin Timothy / CHEM Course: UROP1100, Spring

Machine learning, which is one of the fields in artificial intelligence, is quickly gaining attention after its decisive 4-1 victory against world-renowned Go player, Lee Sedol in 2016. After the success of AlphaGo, “synthesis planning shop (SPLASH)” is created to apply this powerful machine learning into chemistry to benefit chemists and the public in seeking practical and cost-effective pathways for synthesizing all types of molecules, from medicines to polymers. SPLASH is a retrosynthesis algorithm with machine learning capabilities, it utilizes the brain-inspired Artificial Neural Networks (ANN), a machine learning algorithm, called Monte Carlo Tree Search (MCTS). The notations of chemical molecules called SMILES and SMARTS, also python package RDKit are used to let the program comprehend and extract reaction templates from the literature to train the system. The SPLASH system currently has over 4 million reactions and 200 thousand templates but there is a lack of a ranking system to indicate which reactions pathway is the best. This project aims to improve the previously developed SPLASH and propose a new method to find better SPLASH results, through extracting local features and classifying the reaction templates. Support Vector Machine (SVM) is utilized to classify reaction templates according to different reaction types. A new algorithm is proposed to eliminate predicted results that will not react correctly due to selectivity problems, by analyzing the extracted features, which includes functional groups and evaluating the reaction type of reaction template. Through these improvements, SPLASH will hopefully generate more accurate and precise reactions.

5

School of Science Department of Chemistry

Deep Learning in Synthesis Planning Supervisor: SU Haibin / CHEM Student: PHAM Trung Kien / DSCT Course: UROP1100, Fall

In the last few decades, machine learning has significantly emerged itself to become a prominently effective tool in the field of chemistry due to its capability to support traditional hands-on methods of laboratory-based chemical synthesis processes. Integrating the well-known brain-inspired Multilayer Perceptrons (MLP) Artificial Neural Network (ANN) mode, a specific machine learning algorithm Monte Carlo Tree Search (MCTS) and some other cheminformatics-related toolkits and software, a retrosynthesis system is designed and developed to predict stepwise synthesis planning routes for desired target organic molecules. After the training based on 160 molecule structures and nearly 40.000 chemical reactions, the system can consistently predict potential retrosynthetic plans for synthesizing simple linear molecules with accuracy. However, as the system fails to generate proper results for complex molecules with the commercial and feasibility criteria have not been considered yet, further implementation is needed to be accomplished so that the system can provide more comprehensive prediction. Prediction of Inorganic Solar Cell Materials with Double Perovskite Structure Using Machine Learning Approach Supervisor: SU Haibin / CHEM Student: MAO Yun-wen / CHEM Course: UROP2100, Fall

Solar cell has always been one of the focuses of sustainable energy research. However, the fabrication costs of traditional devices remain high. Among many alternatives, perovskite attracts great attention due to its lower costs and higher power conversion efficiency (PCE), which improved from 3.8% to 25.2% within a decade. However, the current best performing perovskites are highly unstable and toxic (contains lead). To overcome these issues, we investigated double perovskite, which was known for higher stability. Kernel Ridge Regression (KRR) Machine Learning (ML) method was utilized to screen 16431 unknown double perovskites and accurately predicted their bandgaps (RMSE = 0.254 eV), stabilities (formation energy RMSE = 2.53 × 10−2 eV/atom), bulk and shear elasticities (RMSE = 2.118 GPa and RMSE= 0.904 GPa., respectively).

Organocatalytic Synthesis of Useful Chiral Molecules Supervisor: SUN Jianwei / CHEM Student: WONG Ting Hei Matthew / CHEM Course: UROP3100, Spring

A series of experiments have been performed in hopes of achieving the asymmetric synthesis of saturated heterocycles catalyzed by chiral phosphoric acid. Substrate design includes a tethered nucleophilic moiety for intramolecular nucleophilic cyclization reaction. Preliminary testing with strong Bronsted acids showed that the reaction is highly efficient, with a short reaction time and high yield. We envision that the reaction should have decent chirality control and an inclusive substrate scope. Currently, we are attempting to find a way to synthesize substrates conveniently and systematically (either through a Friedel-Crafts reaction or direct hydration of a chloride if the former is unviable). Also, the investigation surrounding asymmetric catalyst should begin in the near future.

6

School of Science Division of Life Science Division of Life Science Molecular Regulation of Muscle Stem Cell Quiescence by Non-coding RNAs Supervisor: CHEUNG Tom / LIFS Student: ZHANG Wenxin / BCB Course: UROP4100, Fall

The muscle stem cell (MSC), also called satellite cell, was named based on the location under the basement membrane outside the myofibers. Satellite cells possess remarkable regenerative capacity which allow them to participate in the development and regeneration of the host tissue. Under normal condition, satellite cells remain in quiescence state in the muscle, but it can quickly enter the cell cycle and produce massive myoblasts which fuse to generate muscle fibres once activated by injury. The proliferation and differentiation of MSCs is controlled by multiple factors, including innervation, hormones, and tissue injury. Cytoplasmic polyadenylation element binding protein (CPEB) is one of the modulation protein function during this process. It is a widely distributed RNA-recognition motif that identifies a certain sequence of specific mRNA (CPE) and regulates the translation. The previous study showed the regulatory role of the CPEB family for another gene expression during the stem cell and tissue development. In this project, we try to figure out the role of CPEB1 in satellite cell differentiation and muscle regeneration.

Molecular Regulation of Skeletal Muscle Stem Cell Quiescence and Activation Supervisor: CHEUNG Tom / LIFS Student: BHARATI Ayushi / BIOT Course: UROP1100, Fall

Stem Cells are cells capable of self-renewal and differentiation into different cell types, hence becoming increasingly popular in therapeutic biology. A large subset of adult stem cells, remain in a stage of relative dormancy referred to as quiescence, before returning back into the cell cycle, upon appropriate stimulus. By determining the mechanisms of quiescence and their activation, key findings may be established, providing additional medical and industrial applications. Muscle stem cells are commonly used to understand the mechanism due to the innate stability of muscle cells. Prior to molecular analysis, isolation and identification of satellite cells, using fluorescence assisted cell sorting and their regulatory sequences using ATAC Seq and ATAC See may be carried out. By analysis of both the regulatory elements and key cells, holistic inferences on the mechanism of muscle stem cell quiescence can be drawn.

Molecular Regulation of Skeletal Muscle Stem Cell Quiescence and Activation Supervisor: CHEUNG Tom / LIFS Student: CHEUNG Ka Yan / BCB Course: UROP1100, Summer

This paper aimed to study the post-transcriptional effects of microRNA-653 (miRNA-653) on satellite cells (SCs) using primary myoblasts (PMs) as a study model. Mass spectrometry was performed on miRNA-653 transfected PMs in order to reveal the alterations of gene expression. 5-Ethynyl-2'-deoxyuridine (Edu) labeling was used to quantify PMs proliferation level after miRNA-653 transfection. It was found that miRNA-653 overexpression resulted a depletion in PMs density and an inhibition of PMs proliferation. Mass spectrometry analysis demonstrated that CDK6 level was 17- fold reduced after miRNA-653 transfection. It was postulated that miRNA-653 suppressed PMs proliferation through downregulating CDK6 gene. The inhibitory effects of miRNA-653 in PMs may apply on satellite cells and modulate quiescent satellite cells activation.

7

School of Science Division of Life Science Evolutionary Art at the Fast Track Supervisor: CHOW King Lau / LIFS Student: PARK Yujung / BTGBM Course: UROP2100, Fall

Plasmids play an evolutionarily important role in rapid adaptation to harsh environments especially for prokaryotes such as bacteria. However, when not necessarily needed for the bacteria, these plasmids simply become a replicative burden, and the descendants without the plasmid will have a slight advantage over ones carrying the plasmid. Thus, these plasmids in the absence of the selection pressure are expected to be lost through multiple generations of division, an evolutionary selection process. However, there is lack of information about the process and it is not well understood and studied how and if the plasmid loss progresses over time. In addition, plasmid-mediated antibiotic resistance is one of the major types of antibiotic resistance which has become one of the biggest threats to global health along with the recent rise of superbugs. Thus, it is worth investigating the quantitative measures of the process of antibiotic resistant plasmid loss in bacteria. The present study aims to track the plasmid loss over time in fast growing E. coli strain RR1 transformed with Ampicillin resistant plasmid, pBS II – KS+, under different settings. This will monitor the selection pressure against an extra piece of DNA in a cell when it carries no selective function. The ratio of bacteria with and without the plasmid has been obtained in 24 hours interval under no selection pressure, and the results show that 90% of plasmid loss happens within the 72 hours period.

Evolutionary Art at the Fast Track Supervisor: CHOW King Lau / LIFS Student: SIU Kwong Tai / PHYS Course: UROP1000, Summer

In this study, we are going to determine whether the concentration of chloramphenicol will affect the loss of antibiotics resistant trait of bacteria. Our preliminary result shows that there is a threshold of concentration of chloramphenicol of liquid medium. When the culturing concentration is lower than the threshold, the loss of antibiotics resistance of E.coli becomes significant. On the contrary, when the culturing concentration is higher than the threshold, there is only a very small fraction of E.coli will lose antibiotics resistance. In addition, when the concentration reaches a certain level, the loss of antibiotics resistance property does not appear. We also found that the above behavior could be approximately described by an exponential relation.

Evolutionary Art at the Fast Track Supervisor: CHOW King Lau / LIFS Student: SUN Huiming / SBM Course: UROP1000, Summer

The function-carrying plasmids widely exist in bacteria, not only conferring their host the plasmid-encoded traits (e.g., bacterial resistance against antibiotics) as a selective advantage, but also adding a metabolic burden to the host to express its genes and replicate the DNA. Under conditions without selection pressure where plasmid-free cells outcompete plasmid-carrying ones, the plasmid would be too costly and are expected to gradually go extinct (Wein et al., 2019). Here, we monitor the rate of plasmid loss in the host population under non-selective conditions with several dilution intervals and aim to eventually establish a quantitative relation between the plasmid losing speed and its size. Evolving the plasmid PBSII-KS+ which encodes ampicillin resistance in Escherichia coli strain RR1, it was discovered that the plasmid losing rate is highly correlated with the dilution interval and the losing speed is accelerated about twice (to more than 80% loss in about 24 hours) after shortened the interval to 8 hours. It was also found that the enlarged plasmid size (PBS-F21A3.3p-) slowed down the losing process in E. coli RR1.

8

School of Science Division of Life Science Genetic Identification of Negative Regulator of Bone Morphogenetic Protein Signaling Pathway Negative Regulator Supervisor: CHOW King Lau / LIFS Student: LI Ping Yin / BIBU Course: UROP2100, Fall UROP3100, Spring

This project aims to discover more about the genetic components required for TGF-Beta signaling pathway. The model organism that we use in this experiment is lon-1 C. elegans, which contains a mutation on lon-1 and has a longer body length compared to that of wild type. Last semester, we had conducted a genetic mapping of lon-1 suppressors showing Dpy phenotype. During this semester, I conducted a three-point testcross on wx197, which is located on chromosome II, with KC83(unc-4(e120) rol-1(e91)II; him-5(e1490)V) and try to find out more information about this suppressor. The work elutes to the possibility of extra cellular matrix as key components that augment the function of lon-1 gene.

Genetic Identification of Negative Regulator of Bone Morphogenetic Protein Signaling Pathway Negative Regulator Supervisor: CHOW King Lau / LIFS Student: LI Qian Xin Andreas / BCB Course: UROP1100, Summer

Size is one of the most significant differences across species, and how animals know and control their body size has been a fundamental question in developmental biology. In C. elegans, their body size is largely controlled by the small, male tail abnormal pathway (Sma/Mab pathway) which is under the conserved Bone Morphogenetic Protein (BMP) regulation. Mutations in many components of this pathway result in animals with altered body size. LON-1, a pathogenesis-related (PR)- protein, was found to be transcriptionally repressed by the pathway and when mutated result in a longer animal. LON-1 has also been shown to repress ploidization in hypodermis. However, the mechanism of how LON-1 bring to the changes in size and ploidization remains unclear. Knowing the localization of LON-1 is crucial for understanding its function and, in general, size control in C. elegans. Previously, studies have shown that LON-1 can have physical interaction with the extracellular protein CRM-1, and sequence analysis also shows a putative transmembrane or a secretion signal peptide at the N-terminus. However, there isn’t any direct evidence to show whether LON-1 functions on the membrane or in the extracellular matrix. In this study, LON-1 was fused with green fluorescent protein (GFP) and expressed in Drosophila S2 cell to confirm its cellular localization.

Analysis of Surface Delivery of Epidermal Growth Factor Receptors Supervisor: GUO Yusong / LIFS Student: LEUNG Chak Man / BCB Course: UROP1000, Summer

Due to the current pandemic situation and uncertainty of the community outbreak in Hong Kong, my UROP supervisor professor Guo made a decision in order to keep the social distancing. He sent me 3 different journals (Pacheco-Fernandez et al., 2020) (Tang et al., 2020) (Wang et al., 2020) and asked me to write summaries of them, attending the lab meetings through Zoom to learn advanced knowledge and decision making mindset of the research field. Below are summaries of those journals. Each summary will be described and explained in abstract, approaches and results, as well as the future direction of the research paper.

9

School of Science Division of Life Science Analysis of Surface Delivery of Epidermal Growth Factor Receptors Supervisor: GUO Yusong / LIFS Student: NG Cheuk Hei / BCB Course: UROP2100, Summer

The rapid spread of a novel and highly pathogenic coronavirus (SARS-CoV2) has caused a serious global public health emergency in 2020. A densely glycosylated spike (S) protein enriched on the surface of SARS-CoV2 is the key player that promotes viral entry into cells. The S protein is a trimeric protein composed of the S1 and S2 subunits. A subdomain of S1, referred to as the receptor-binding domain, directly binds a cell surface receptor, angiotensin-converting enzyme 2 (ACE2). This interaction is critical for viral attachment. To develop effective therapeutic strategies, my project focuses on studying the interaction between ACE2 and CoV2-RBD, an essential step for viral attachment.

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

Due to the pandemic condition in Hong Kong. I was in mainland China during the semester. Dr. Guo asked me to write a summary of paper published in Science (Mukhopadhyay S and Linstedt AD, Science 2012) entitled “Manganese blocks intracellular trafficking of Shiga toxin and protects against Shiga toxicosis”. Shiga toxin (STx)–producing bacteria cause more than a million deaths each year and have no definitive treatment. Antibiotic treatment may cause STx release from the dying bacteria in the human body, causing life-threatening complications. This paper found that being exposed to certain concentration of Mn2+ protects cells from STx toxicity by specifically intervening the biological function of GPP130, a protein involved in the traffic of STx from early endosome to Golgi and helps STx avoid the fate of being transported to and degraded in lysosomes. In addition, the authors showed that Mn2+ was proved to be effective in protecting cells or mice in vivo against STx-induced toxicity. Thus Mn2+ may be a cheap and convenient treatment for STx infection.

Investigating Transcription Elongation Dynamics at a Single-Molecule Level Supervisor: ISHIBASHI Toyotaka / LIFS Student: LEONG Kin Nam / BCB Course: UROP1000, Summer

Previous research has shown that MEF2D (Myocyte-specific enhancer factor 2D) protein enhances polymerase transcription activity, however, a quantitative measurement of its effects has not been done. This project aims to investigate the quantitative effects of MEF2D on the dynamics of RNA polymerase transcription at a single molecule level, using optical tweezers. Results are unavailable at the time this report was written since major obstacles were encountered during control data recording, research on this topic will continue. This report focuses on the explanation of the experiment and the above-mentioned obstacles with, if any yet, probable solutions.

10

School of Science Division of Life Science Understanding the Key Elements of Nucleosome/Chromatin Dynamics Supervisor: ISHIBASHI Toyotaka / LIFS Student: YUK Tin Cheung / BIBU Course: UROP1100, Spring

H2A.P is a type of putative and small DNA binding histone variant that can be found in both mice and human beings. The histone variant shows a significant difference from the canonical histone H2A in terms of localization and amino acid sequence structure. It was found out that the H2A.P has less than 50% amino acid identity with canonical histone H2A. Meanwhile, some studies even revealed that the H2A.P differ from canonical histone to an extent that can influence chromatin function. For instance, H2A.P has a truncated C-terminal docking domain which is normally used to maintain the structural properties for the nucleosome. In terms of localization, level of H2A.P is found to be exceptionally and exclusively high during the stage of male’s germ cell development whereas the canonical histones should be found universally in different cell types processing DNA.

Deciphering Epigenetic Changes in Cancer -**UROP-EPIC program** Supervisor: LEUNG Danny Chi Yeu / LIFS Student: KRISHNAMOORTHY Suhas / BIBU Course: UROP1100, Summer

The field of epigenetics has rapidly developed over the past 30 years through technological advances and better understanding of key molecular mechanisms in epigenetic control. This project focuses on understanding epigenetic pathways, how they control expression and the recent developments in the field in order to gain insight into the impact and processes of epigenetic research through a comprehensive literature review followed by an analysis of publicly available epigenomic data by utilizing several bioinformatic tools on Galaxy, a web-based platform for biological research. The histone modification ChIP-seq data processing pipeline from ENCODE is applied to characterize the cisregulatory features of H3K27ac enrichment in mouse embryonic stem cells.

Deciphering Epigenetic Changes in Cancer -**UROP-EPIC program** Supervisor: LEUNG Danny Chi Yeu / LIFS Student: SHAH Yashvi / SSCI Course: UROP1100, Fall The final phenotype is not only determined by the nucleotide sequence, but also by the epigenetic changes on the double helix, more specifically the DNA methylation that occurs on cytosine residues. This paper discusses the molecular biology methods to understand the effect of the nucleotide sequence and the epigenetic changes on the DNA on the phenotype, and the future direction of these experiments.

11

School of Science Division of Life Science Proteomics Approach to Decipher Gravity Signalling in a Flowering Model Plant Supervisor: LI Ning / LIFS Student: SIU Kwong Tai / PHYS Course: UROP1100, Spring

The research on protein structure and protein-protein interactions is important to study their biological functions. There are multiple approaches to understand protein structure and interactions in modern biology, including magnetic resonance imaging (MRI) and X-ray crystallography. However, these traditional methods have rigorous requirements on the preparation of protein samples. The cross-linking mass spectrometry provides a new approach to interactomics and structural biology. Compare to traditional methods, cross-linking mass spectrometry has advantages on costs, sensitivity and low requirements on protein samples. This article will introduce the basic principle of cross-linking mass spectrometry and its applications, including some popular cleavable and enrichable cross-linkers and software on identification and quantification of cross-linked peptides.

Anticancer Drug Resistant Mechanisms Supervisor: LIANG Chun / LIFS Student: CHIU Yui Hei / SSCI Course: UROP1000, Summer

M2 is a potential anti-cancer drug candidate previously discovered in Dr Liang Chun’s laboratory, which is originated from traditional Chinese herbal medicine. 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 problem which many drugs face – Drug resistance. Hence, this progress report devotes itself into studying the anticancer drug resistant mechanisms involved by characterising a resistance population- R8, a cell line developed from Hela Cell line. We found out that R8 population has a very distinctive morphological and chemically compared to the Hela cell line, we believe that the differences are responsible for the aggressive and invasive behaviour. On top of that, we isolated R8 single cell clones and characterized for M2 resistance. We found out that R8 clones display higher survivability and recoverability in the presence of drug compared to parental hela cell lines.

Anticancer Drug Resistant Mechanisms Supervisor: LIANG Chun / LIFS Student: HO Phei Ting / BIOT Course: UROP1000, Summer

The aim of this report is to further enhance self-thorough understanding on the fundamental steps of cell culture, cell seeding, cell viability assay and glucose uptake assay on HepG2 cells. Furthermore, different applications of drugs (Traditional Chinese medicine) are tested on insulin resistance (IR) HepG2 cells in comparison with a positive control of Western medicine (Metformin) on diabetic subjects. In particular the 3 herbs: Momordica charantia, Astragalus membranaceus, and Pericarpium granati. HepG2 cells were exposed to high concentration of insulin and glucose to mimic type 2 diabetic models in HepG2 cells. The insulin resistance (IR) HepG2 cells are then treated with the 3 Chinese herbs of different concentrations and analyzed the glucose uptake and cell viability. Future self-goal is to analyze specific protein level expression (GLUT4, IRS-1, pAkt, NF-κB, AMPK) using Western Blot to further utilize the optimum traditional Chinese herb effect on IR HepG2 cells.

12

School of Science Division of Life Science

DNA Replication-initiation Proteins in Budding Yeast Supervisor: LIANG Chun / LIFS Student: ANSAR Aneeta / BCB WONG Tsun Chiu / SSCI Course: UROP1000, Summer UROP1000, Summer

Noc3p is a key protein that is required for DNA replication licensing. It is needed for the assembly of various DNA replication-initiation proteins (DRIPs) to initiate DNA replication. Based on previous studies done in our lab, it is known that Noc3p and ORC (Origin Replication Complex) subunits self-interact and form dimers during the process of pre-RC formation. However, the mechanism of such self-interaction remains unclear. To further elucidate the mechanism of Noc3p self-interaction, in this study, NOC3-FRB cells had been used to deplete free Noc3p from the yeast cell nucleus and to determine the impact of depletion on Orc6p self-interaction.

DNA Replication-initiation Proteins in Budding Yeast Supervisor: LIANG Chun / LIFS Student: KWAN Chi Hang / BIBU NG Cheuk Hei / BCB Course: UROP1100, Fall UROP1100, Fall

The cell cycle assembly and disassembly of DNA replication-initiation proteins is a highly regulated process. The origin recognition complex (Orc1-6p) binds and marks replication origins to facilitate Cdc6p dependent Minichromosome Maintenance (MCM) complex association during the M-to-G1 transition. Noc3p has been established as an essential DNA replication initiation protein. Despite substantial gains in our understanding of this process there are still several mechanisms that have yet to be fully elucidated. Recently, we have identified that Noc3p like ORC has cell cycle dependent dimerization. In this investigation we report that Noc3p dimerization is independent of ORC dimerization and Cdc6p.

Potential Cancer Drug Identification and Characterization Supervisor: LIANG Chun / LIFS Student: ADELINE Priscilla / BIBU KWOK Nicholas Tze Yin / BIBU Course: UROP1000, Summer UROP1000, Summer

Cancer is highly diverse in nature, with high mutation rate and treatment resistance, leading to the massive demand in research for effective novel cancer drugs. Thus, an effective way to screen for novel target drugs is crucial to drug development optimization and further cancer biology discovery. In this study, a drug candidate, namely IF5, has shown its cancer inhibition potential through the screening process. Using flow cytometry for cell cycle analysis, the result has also shown IF5 characteristics on disrupting certain cellular pathways in the G1 phase to S phase and leading to cytotoxic effect on HeLa, predominantly once it entered S phase.

13

School of Science Division of Life Science

Molecular Regulation of Axon Regeneration Supervisor: LIU Kai / LIFS Student: CAO Jiaming / BCB Course: UROP2100, Fall UROP3100, Summer

The human nervous system consists of central nervous system (CNS) and peripheral nervous system (PNS). In adult mammals, PNS axons can regenerate after injury while CNS axons do not have the ability to regenerate after lesion. This has been a serious challenge in clinical neuroscience for many decades, since there is no cure for spinal cord injury still. Understanding CNS axon regeneration is an important step that in the future could lead to treatments for spinal cord injury, stroke, ALS and also other neurodegenerative disorders. There are both extrinsic factors and intrinsic factors for the regeneration failure of adult mammalian CNS axons, and the molecular mechanisms behind axon regeneration is still under research currently.

Molecular Regulation of Axon Regeneration Supervisor: LIU Kai / LIFS Student: GAO Zhouyang / BCB Course: UROP1100, Fall UROP2100, Spring

The regeneration failure of axons in the mature mammalian central nervous system (CNS) after injury is commonly observed. The inability of lesioned CNS axon to regeneration can result in permanent disabilities. The regeneration failure of axon after injury to CNS contrast with the robust regenerative ability of the peripheral nervous system (PNS) axon after injury. The mechanism behind this phenomenon draws the attention of considerable studies over the years. In this review, a number of common models of studying axon regeneration mechanisms are examined. Furthermore, the injury response and the intrinsic factors on axon regeneration failure at the cellular and molecular level will be discussed.

The Rules of Packaging Fat in Cells Supervisor: MAK Ho Yi / LIFS Student: MO Hok Chun / BCB Course: UROP2100, Spring SEIP-1, the seipin ortholog in C. elegans, was previously reported to be enriched in membrane structures associate with a subset of lipid droplets in nematodes’ intestinal cells. Further studies have revealed the presence of similar SEIP-1 positive structures in oocytes and early embryos. In order to determine the abundance and fate of SEIP-1 positive structures prior to and immediately after fertilization, I generated a transgenic worm strain that express a fusion protein of SEIP-1 with the photoconvertible fluorescent protein, mKikGR. Here, I report the preliminary results from live imaging experiments.

14

School of Science Division of Life Science

A Study of Zebrafish Behavior with respect to Water Temperature Gradients Supervisor: MILLER Andrew Leitch / LIFS Co-Supervisor: HO Sarah Elizabeth / LIFS Student: MA Chengkun / BCB Course: UROP1100, Spring

Zebrafish is a powerful model system in biomedical research. This project aims to explore the possibility of using zebrafish to screen anti-emesis drugs with its thermoregulatory behavior. In this semester, this project focuses on literature review and the experiment design. This report will first discuss the advantages of using zebrafish in research which are the reasons. Then, this report talks about the nausea and hypothermia induced by motion sickness in humans which is followed by the behavioral thermoregulation in zebrafish. Finally, an experiment testing the response of zebrafish to emesis-inducing drugs is designed based on previous experiments in zebrafish thermoregulatory behavior. This experiment is expected to be conducted in the following semester under this project.

Characterization of Novel Cell Cycle Regulators in Cancer Cells Supervisor: POON Randy Yat Choi / LIFS Student: CHEUNG Lo Sha / BCB Course: UROP1000, Summer As a member of MYC-family, MYCN gene amplification has been proved that related to multiple human tumors, including a severe form of neuroblastoma. In this research project, by deploying CRISPR-Cas9 technology, different sequences are targeted, causing specific sites of double-strand break (DSB) on MYCN gene, inducing deactivation of NMYC genes. Later, the protein expression level of N-Myc and DNA damage responses on studied cell lines were examined. A MYCN CRISPR targets exon 2’s open reading frame has been proved that can turn down exogenous MYCN overexpression. Since many cancers are related to MYCN irregulated expression, the study of MYCN CRISPRs inside can provides insights for therapeutic development targeting MYCN gene.

Characterization of Novel Cell Cycle Regulators in Cancer Cells Supervisor: POON Randy Yat Choi / LIFS Student: INTAN Angelica Grace / BIOT Course: UROP1100, Spring

Cancer cells are mutated such that they are able to proliferate indefinitely. The mutations usually occur on cell-cycle regulators such as the Cyclin E1 and E2. In order to develop treatments for cancer, a deeper understanding of cancer cell cycle regulators is necessary. In previous studies, it was found that Cyclin E1 and Cyclin E2 are functionally redundant; a knockout of either one will produce normal phenotypes. In this study, Cyclin E2 knockout genes are introduced into HeLa cells in order to determine the role that Cyclin E2 plays in cancer cell proliferation. Due to limited time, the experiments performed failed to accomplish this goal, but through these experiments, two reagents, the cyclin E2 antibody from Santa Cruz Biotechnology and CRISPR-knockout Cyclin E2 gene, were found to be faulty and discarded such that future experiments performed in this lab would not be affected by the dysfunctional reagents.

15

School of Science Division of Life Science

Characterization of Novel Cell Cycle Regulators in Cancer Cells Supervisor: POON Randy Yat Choi / LIFS Student: JUNG Hui Jeong / BIOT Course: UROP1100, Fall

The iron-sulfur cluster assembly 1 and 2 genes, or ISCA1 and 2, are mitochondrial proteins that are involved in the assembly of iron-sulfur protein clusters, which play a role in electron-transfer reactions. The Multiple Mitochondrial Dysfunctions Syndrome is associated with the mutation of ISCA1 and 2, and this disease leads to psychomotor development loss and neurologic deterioration in early infancy, thereby contributing to death in early childhood. In this study we attempted to construct the complete knockout cell lines using the CRISPR approach to identify the possible role(s) of these two genes in cell survival, cell proliferation and cell cycle progression.

CRISPR/Cas9 Analysis of Essential Genes Supervisor: POON Randy Yat Choi / LIFS Co-Supervisor: MA Hoi Tang / LIFS Student: KORANI Lavisha Ramesh Kumar / BIBU Course: UROP1000, Summer

Aurora A kinase is a key regulator of mitosis. Both the polo-like kinase PLK1 and Bora are conserved Aurora A activators. Together, these proteins are important in the DNA damage checkpoint which prevents cells from entering mitosis after DNA damage. Using technologies including CRISPR-mediated depletion, we have characterized the effects of ionizing radiation on Aurora A. The amount of phosphorylated Aurora A during G2 phase was shown to be reduced after DNA damage. Progress has also been made in generating CRISPR-mediated disruption of Bora for studying the role of Bora in DNA damage checkpoint.

Cloning and Characterization of Novel Cytoskeletal Regulators Supervisor: QI Robert Zhong / LIFS Student: CHEUNG Yan Tung / SSCI Course: UROP1000, Summer

The cytoskeleton is an inevitable interlinking network for regulating cellular processes. It is crucial to understand the cytoskeleton-related proteins as they control the mechanisms of the cytoskeleton. The protein GAS2L1 contains CH and GAR domain that binds to microtubules and actin filaments respectively. Previous researches showed that autoinhibition exists in bindings among the two domains, while the structural detail remains unknown. Therefore, we seek to find out the suitable condition of GAS2L1 for further biophysical assay. Molecular cloning techniques were first used to clone our desired GAS2L1 fragments into bacterial expression plasmids, and the protein was later purified from bacterial culture. The GAS2L1 fragment containing both CH and GAR domain (residue 1-284) was successfully cloned and purified in this project.

16

School of Science Division of Life Science

Cloning and Characterization of Novel Cytoskeletal Regulators Supervisor: QI Robert Zhong / LIFS Student: LAU Yee Hin / BIBU Course: UROP1100, Summer

The γ-tublin ring complexes (γ-TuRC) is a major regulator of microtubules, which forms a part of the cytoskeleton. While the main protein constituents of γ-TuRC have been identified, not all of its members have their function fully elucidated, in particular GCP 4, GCP5 and GCP6. This project has taken a preliminary step in investigating their function by using CRISPR Cas9 to tag the proteins of interest with the auxin inducible degron (AID) system to reversibly deplete them. This project focuses on GCP5 and successfully generated 3 sgRNA clones that will enable GCP5 to be conjugated with the AID tag at the N-terminus.

Computational Study of Long Noncoding RNAs in Cancer Supervisor: WANG Jiguang / LIFS Student: LEE Jooran / BIEN Course: UROP1000, Summer

To identify upregulated and downregulated in the recurrent samples after the treatment with temozolomide (TMZ), the differential analysis using DESeq2 has been performed and several potential long non-coding RNAs (lncRNAs) were identified. As the aggressive phenotype of the gene is related to shorter overall survival and earlier progression, the survival plots using sample information have been drawn and it was revealed that the age is associated with the worse survival. Lastly, it was also proved that it is impossible to separate initial and recurrent GBM samples using the lncRNA expression profile. In doing so, the consensus clustering and the heatmap were adopted to differentiate the samples.

Study of Blood Cell Development using Zebrafish Model Supervisor: WEN Zilong / LIFS Student: LI Yunbo / BCB Course: UROP2100, Summer

mstf1 is a specifically expressed gene identified in metaphocyte, a distinct myeloid-like cell population that shares high similarities with Langerhans cell (LC) yet originates from a different germ layer. To characterize the expression pattern of mstf1 and to label metaphocytes in vivo, a reporter line Tg(mstf1:GFP) was generated. Now, we have obtained several Tg(mstf1:GFP) founders and designed various experiments to evaluate the labeling efficiency of these lines and investigate the distribution of mstf1-expressing cells. We will also examine the function of mstf1 by ectopically expressing it in other cell populations. The construction of plasmids for the ectopic expression of mstf1 is in progress. Besides the lab work, mechanisms of two highlighted techniques-- fluorescence-activated cell sorting and tissue cryosection were also introduced.

17

School of Science Division of Life Science

G Proteins and their Regulators in Cancer Biology Supervisor: WONG Yung Hou / LIFS Student: LAM Cheuk Him / BIBU Course: UROP1100, Summer

It has been discovered that condition medium produced by U87MG astrocytoma cells through activation of its Gq coupled receptor with Bradykinin (BK) can induce increased level of phosphorylated STAT3 production in SKNMC neuronal cells, meaning that these astrocytes are capable of releasing substances which increase neuron’s ability of survival through activating the STAT3 pathway responsible for self-renewal and proliferation. To consolidate that this form of astrocytic-neuronal communication is in fact a universal process for human brain cells, the experiment was repeated by using another glial cell line, the 1321N1 cells, to produce the condition medium that was applied to SKNMC cells. Observed from the western blotting results, STAT3 activation in SKNMC cells can be increased by neurotransmitters secreted by 1321N1 under influence of Bradykinin, however, the results from SKNMC cells treated with U87MG condition medium (CM) were not significant during my study.

Interdisciplinary Bioengineering: Microfluidic Device for Cell Manipulation/Culture and Molecular Biology Supervisor: WU Angela Ruohao / LIFS Student: KUSUMA Alvin Riva / BIOT Course: UROP1100, Fall

Microfluidics is defined as a technology to manipulate fluids at a micro scale. Microfluidics has developed in its methods and efficiency in the past 30 years, especially in the realm of life science studies. The scope of microfluidics in biological studies can range from microbiology, drug testing and assay, cell cycle studies, and recombinant DNA technology. A prominent and commonly used types of microfluidic in life science study is PDMS-based device, due to its inert nature, minimizing interaction with the biological sample and assay performed, and due to its transparence. PDMS is also versatile and flexible on its usage and development.

18

School of Science Division of Life Science

Interdisciplinary Bioengineering: Microfluidic Device for Cell Manipulation/Culture and Molecular Biology Supervisor: WU Angela Ruohao / LIFS Student: LEUNG Ziuwin / BIOT Course: UROP1100, Fall

Sea urchins are said to be resilient to ocean acidification. To fully understand the epigenetic effects of the current environmental changes on sea urchins, single-cell analysis is used to compare and identify select genes that are responsible for its resilient characteristics, however, its standing genetic variation poses as a challenge. To minimize the genetic variability within a population, monozygotic sea urchin twins can be used as an alternative as the difference in genetic expression can be correlated to exposure of different environmental factors and not due to its genetic variability. However, conventional methods for embryo twinning require labor-intensive and high precision cutting by skilled technicians, limiting the methods throughput and its applicability for single-cell analysis. In this project, we proposed a microfluidic lab-on-a-chip prototype for studying marine organism tolerance to ocean acidification through the automation of embryo splitting of sea urchin embryos into two monozygotic twins. The project consists of two parts: (1) the alignment of 2-cell stage cleavage and knife with dielectrophoresis (DEP) and (2) the splitting of the embryo. Through numerical simulation of different electrode geometries, electric fields, and DEP forces, a proper design of electrodes and a set of DEP parameters were derived and tested to study the embryo's response to DEP. In part 2 of the project, a microknife design, adapted from a microguillotine device, is integrated into the prototype design once alignment in part 1 has been achieved. This progress report focuses on part 1 of the project to better understand how DEP can be used to orientate the embryos.

Human Complex Disease Genomics and Bioinformatics Supervisor: XUE Hong / LIFS Co-Supervisor: LIANG Chun / LIFS Student: IU Wing Ting / BIBU Course: UROP1100, Summer

The genetic causes of different human complex diseases are one of the most investigated research topics. For schizophrenia, mutations associated with the GABARB2 gene, which encodes the β2 subunit of the GABAA receptor in the central nervous system, is one of the possible genetic causes discovered recently. This project aims to study more about the genetic causes of schizophrenia and other complex diseases utilizing a DNA library. The AluScan method is used to facilitate gene amplification for better identification of genetic polymorphisms. To obtain suitable DNA fragments for the DNA library, a series of experiments have been done: Alu PCR, fragmentation, end repairing, adenylation, ligation, and KOD PCR.

Human Complex Disease Genomics and Bioinformatics Supervisor: XUE Hong / LIFS Co-Supervisor: LIANG Chun / LIFS Student: KAN Chloe / BIBU Course: UROP1100, Spring UROP2100, Summer

GABA is the main inhibitory neurotransmitter that binds to GABAA receptors. GABAA receptors have been found to be related to complex disorders such as epilepsy and schizophrenia, thus it has been the main focus in drug development. The structure of GABAA receptors allows for multiple binding sites, including binding of the GABA and various pharmacological drugs, such as benzodiazepines. However, these drugs may have adverse side effects; hence recent studies aim to identify naturally occurring GABAA receptor ligands from traditional medicinal herbs. This study aims to investigate the structure of the GABAA receptor, particularly the α1 subunit, in order to further explore the binding mechanism of drugs to the receptor.

19

School of Science Division of Life Science

Receptor based Drug Development from Chinese Herbal Medicine Supervisor: XUE Hong / LIFS Student: CHEUNG Kai Chit / BCB MA Chengkun / SSCI Course: UROP1100, Fall UROP1100, Fall

Neurotransmitters are typically classified as excitatory and inhibitory ones. This lab aims to study the structure, functions of an inhibitory neurotransmitter and its receptor – GABA and GABAA receptor, as well as the treatment of its defect. The research combines this receptor with drugs from Chinese herbal medicine and looks for better solutions to related diseases. During the UROP, we assisted Ph.D. candidates in our lab with daily lab routines and experiments. To find the relationship between GABAA receptors and related disease such as schizophrenia, a knockout mice model was proposed. We performed mice genotyping and DNA library construction which are parts of the ongoing research in the lab. This gave us valuable chances to practice our experiment technique in DNA extraction, PCR, gel electrophoresis, etc.

Receptor based Drug Development from Chinese Herbal Medicine Supervisor: XUE Hong / LIFS Student: CHEUNG Kai Chit / BCB Course: UROP2100, Spring

Human complex diseases have long been in specific scientific interest. With the advancement of sequencing techniques, Next-generation sequencing (NGS) can be adopted to carry out genetic studies on genes that are potentially related to human complex diseases. To prepare the samples for NGS, DNA library construction can be used which comprises of multiple steps, including Alu PCR, fragmentation, end repairing, adenylation, adapter ligation and indexes insertion to identify and generate DNA fragments for sequencing. In this experiment, 4 Alu products were used for DNA library construction and 3 sets of libraries with concentration between 13.5 ng/μl to 17.2 ng/μl were obtained. The products will be used for further analysis to locate mutations with specific pharmaceutical interest.

Receptor based Drug Development from Chinese Herbal Medicine Supervisor: XUE Hong / LIFS Student: TSUI Shuk Man / BCB Course: UROP2100, Fall UROP3100, Spring UROP4100, Summer

Multiple single-nucleotide polymorphisms had been identified in GABAA receptors β2 subunit gene (GABRB2) to be associated with schizophrenia (SCZ). Gabrb2 knock-out (KO) mice, which exhibited extensive similarity to the phenotypes of human SCZ patients, was developed to investigate the role of GABRB2 in the ontology of SCZ. This study aimed to identify the linkage between the loss of GABRB2 and SCZ phenotypes. Transcriptomic analysis was carried out to identify differentially expressed genes (DEGs) in the Gabrb2 KO mice model compared to their wildtype (WT) as a control group. The significantly differentially expressed genes were found to be associated with cell proliferation, neuroinflammation, lipid metabolisms, and oxidative stress response. These suggest a disruption in hippocampal adult neurogenesis and accumulation of oxidative stress induced by neuroinflammation.

20

School of Science Division of Life Science

Molecular Mechanism of Cell Competition and Tumorigenesis Supervisor: YAN Yan / LIFS Student: LI Qian Xin Andreas / BCB Course: UROP1100, Spring

Cell competition is an intrinsic tissue optimizing system in which viable cells with lower fitness are eliminated by its neighboring fitter cells. This process was first described in Drosophila bearing the mutant cells which have a slower growth rate, but was later found to be conserved over species. Fitness indicators other than growth rate, such as cell polarity, were also reported. The mechanism of cell competition has gained much interest as it has shown to be a potential tumor suppressing system. In this report, we try to understand the molecular mechanism of Scrib induced cell competition by using single cell RNA sequencing. Here we present a list of genes that potentially involved in cell competition, and this provides some targets for future research.

Molecular Mechanism of Cell Competition and Tumorigenesis Supervisor: YAN Yan / LIFS Student: LI Xingzhi / BTGBM Course: UROP1100, Fall

The c-Jun N-terminal kinases (JNKs), together with ERK signaling activities, are 2 pathways belonging to mitogenactivated protein kinase (MAPK) family that contribute to the control of differentiation, apoptosis, and proliferation. Under the context of oncogenic Ras, through the activation of JNK signaling pathway, mutation of different apicobasal polarity genes such as a group of neoplastic tumor suppressor gene(nTSG) including scrib, lgl, and dlg promotes both overgrowth and metastatic behavior of the tumor in the model animal of the experiment, Drosophila. Since the screening list from RNA sequencing that correlated with JNK reporter's expression pattern has been established, our work aimed to identify more genes whose knockout leads to enlarged tumors in Drosophila imaginary wing disk as JNK does.

Molecular Mechanism of Cell Competition and Tumorigenesis Supervisor: YAN Yan / LIFS Student: SONG Weixuan / BCB Course: UROP1100, Spring UROP2100, Summer

This summer term, I continued with my experiments in the spring semester, which aimed to study the growth suppressed phenotype in the early stage of TSGs mutant drosophila imaginal wing disc tumors. Last semester, I focused on the relationship between insulin level and imaginal wing disc tumor suppression. And this semester I started to look into the relationship between Drosophila tumor-necrosis factor alpha (TNF-α), which is Eiger (egr) protein in this case and the tumor growth in the early stage. Eiger (egr) protein is produced by fat body, the adipose-like tissue in Drosophila. We are also doing some screening to try to study the mechanism of Eiger protein amount control in fat body.

21

School of Science Department of Mathematics

Department of Mathematics Low-Rank and Sparsity Reconstruction in Data Science Supervisor: CAI Jianfeng / MATH Co-Supervisor: YE Guibo / MATH Student: CHAN Tsz Ho / DSCT Z HANG Liangwei / DSCT Course: UROP2100, Fall UROP2100, Fall

In many studies about neural nets, neural network training always relies on our own ability to find a good minimizer of loss function. While certain network architecture can produce a loss function that trains faster and easier, well-defined training parameters produce minimizers that generalize better. Based on that, we address the issue whether sharpness of minimizers correlated well with generalization error. To visualize the loss of trained neural networks, 1-Dimensional Linear Interpolation is used to compare loss functions with two different parameters in same plot. Training batch-size is controlled while other conditions remain same to get the side-by-side comparison about the sharpness. Besides, we change some parameters to verify the relationship between sharpness and generalization. The result with current setting shows contradiction with the widely accepted knowledge and needed further exploration.

Low-Rank and Sparsity Reconstruction in Data Science Supervisor: CAI Jianfeng / MATH Co-Supervisor: YE Guibo / MATH Student: CHEUNG Cheuk Lun Alan / MATH-IRE Course: UROP1100, Summer

Higher-order tensors are applicable in many academic fields, including statistics, signal processing, and scientific computing. In working with high order tensor, it is more efficient to decompose tensor into lower rank components for analysis. Therefore algorithms designed for this purpose are important. In this paper, we would look at the algorithm using Riemannian Trust Region Scheme. The trust-region method is an iterative method designed to minimize the cost function, which is a real-calued function measuring the difference between our approximation and the real tensor. This paper would focus on the steps of this algorithms and the computation of gradient and Hessian of the cost function. To simplify the explanation, third-order tensor would be used as examples. Low-Rank and Sparsity Reconstruction in Data Science Supervisor: CAI Jianfeng / MATH Co-Supervisor: YE Guibo / MATH Student: GUO Linyin / ACCT Course: UROP1100, Summer

This article is the final report of UROP1100, within which the author focused on sparsity in data science and machine learning. By using a case of the European migration flow, this article explicitly shows the noise in data collected in real-world problems. The article continues to illustrate the ideas and models of optimal transport, regularized optimal transport, and inverse optimal transport. Finally, some preliminary conclusions and possible future progress are mentioned to point out the failure and further expectations of this study. Noticeably, this article mainly aims to detail the author’s knowledge and understanding of the aforementioned areas in two months. Hopefully, this article will be a useful assessment of the project.

22

School of Science Department of Mathematics

Low-Rank and Sparsity Reconstruction in Data Science Supervisor: CAI Jianfeng / MATH Co-Supervisor: YE Guibo / MATH Student: LIU Xinyu / MAEC Course: UROP1100, Fall

In this UROP project, I am discovering a framework for solving non-convex optimization problems under certain dual conditions. When these conditions are satisfied, the non-convex optimization problems can be converted to convex optimization problems, which can be solved in an easier, more regular procedure. In reality, many matrix factorization problems, which are widely applied in topics of machine learning such as predictions and recommendation systems, satisfy these conditions and therefore can be simplified by this framework. In this report, I will firstly introduce some basic definitions in optimization, then one of the core idea behind the problem conversion—Low-Rank Approximation. In the end, the general mathematical ideas behind framework will be presented.

Low-Rank and Sparsity Reconstruction in Data Science Supervisor: CAI Jianfeng / MATH Co-Supervisor: YE Guibo / MATH Student: LYU Hanfang / MATH-AM WU Ruirui / MATH Course: UROP1100, Fall UROP1100, Fall

In UROP1100 this semester, we were mainly learning the mathematics behind the Variational Autoencoder (VAE) and trying to explain it in our own way. We basically learned it from a tutorial website and some papers, and wrote the report together. During the whole progress, under the guidance of Professor Cai, we discussed with each other and learned a lot.

In this report, WU Ruirui wrote all the things in Introduction section and Setting Up the Goal in the General Method section, which explained what is VAE and what are we going to do. Meanwhile, I (LYU Hanfang) was in charge of all the other things including the Appendix, mainly calculating and explain some tricks on how to implement VAE.

Owing to the lack of basic knowledge in Machine Learning, it's a pity that we don't have time to learn it and are not able to implement the VAE. Please point out our unprofessional mistakes or misunderstandings if there is any. Thank you!

Low-Rank and Sparsity Reconstruction in Data Science Supervisor: CAI Jianfeng / MATH Co-Supervisor: YE Guibo / MATH Student: SUN Yuran / MATH-SFM Course: UROP2100, Fall

Last time when we study the low-rank recovery, we found that the accuracy of the initialization part was a bit low so we aim to find the optimal initialization (with highest accuracy) when recovering a low-rank matrix from phaseless linear projections of each of its columns this time. We study the performance of the spectral method in the highdimensional limit first. We also analyze some examples and aimed to find the optimal preprocessing functions and the optimal performance curves.

23

School of Science Department of Mathematics

Low-Rank and Sparsity Reconstruction in Data Science Supervisor: CAI Jianfeng / MATH Co-Supervisor: YE Guibo / MATH Student: ZENG Ziming / ECOF Course: UROP1100, Summer

Nowadays, data can be everywhere. Data matching has become a critical problem with different kinds of applications in many aspects of the real world. This report provides a thorough analysis of Inverse Optimal Transport through making exposition and comparing Optimal Transport, Regularized Optimal Transport and Robust Inverse Optimal Transport. In the introduction, the author makes an elucidation on nowadays’ data application and the importance of data matching. Then the author introduces Optimal Transport, Regularized Optimal Transport and Robust Inverse Optimal Transport briefly. In the case study, the author solves a sample problem to predict the total ride duration of taxi trips in New York City, presenting how the Inverse Optimal Transport work in Python programming in details.

Statistical analysis of competition data Supervisor: CHEN Kani / MATH Student: YANG Yihan / RMBI Course: UROP1100, Fall UROP2100, Spring

The project of “Statistical analysis of competition data” in Spring 2020 is the continuous research of Fall 2020 which focuses on the new feature developments as well as the final performance evaluation. The new features are related to players past performance and will work together with the previous features which mainly focus only on the current match information. And the final performance of each model is the cumulative return during the year of 2018 whose calculation is based on the expected return on the predicted betting choice given the result of each games.

Statistical analysis of competition data Supervisor: CHEN Kani / MATH Student: ZHOU Jingan / MATH Course: UROP1100, Fall

Analyze on NBA game outcomes attracts attention among sports analysts due to the availability of a huge amount of sports data. While we can usually generate a decent accuracy using betting odds to make prediction, we are interested in if we can predict NBA game outcomes purely by historical game statistics but with a higher accuracy. We utilize a dataset containing game statistics in the regular season of year 2016 to 2017. Models including logistics regression, random forest, support vector machine, Naïve Bayes, extreme gradient boosting (XGB) are attempted, with emphasize on performing hyperparameter tuning on XGB. Preliminary XGB model achieves 66.5% accuracy with team location as the most important feature. Possible feature engineering methods will also be discussed.

24

School of Science Department of Mathematics

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: CHAN Ka Man / SSCI WU Ruirui / MATH-PMA Course: UROP1100, Summer UROP1100, Summer

In this report, we will present the paper THE HEAT EQUATION SHRINKING CONVEX PLANE CURVES by Gage and Hamilton and paper CURVE SHORTENING MAKES CONVEX CURVES CIRCULAR by Gage. These are two fundamental papers on curve shortening flow which show that a smooth convex curve will shrink to a point circularly under heat equation. We mainly focus on how the proof of main theorem goes in them, and complete some details omitted by authors.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: CHEN Zijun / MATH-PM Course: UROP1100, Fall UROP2100, Summer

The report mainly uses the maximum principle to research on the compact 3-manifold. Assume the initial Ricci and scalar curvature is positive. The report discusses the estimate regard with the scalar curvature, the covariant derivativeof scalar curvature and the limit of 𝑅𝑅𝑚𝑚𝑎𝑎𝑥𝑥 𝑅𝑅𝑚𝑚𝑚𝑚𝑚𝑚. The final goal of the report is to show that any compact 3-manifold will admits a metric of constant positive curvature if the manifold admits a strictly positive Ricci curvature. The report first discusses with the unnormalized condition.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: HUANG Meng / MATH-PMA Course: UROP1000, Summer

This UROP project centered around two papers about curve shortening flow written by M. Gage. The goal of the two papers is, in short, to prove that when a curve in 2D evolves according to the equation ∂F/∂t = nκ, it will 1) converge to apoint, 2) converge to a unit circle when normalized in size. This report will explain some selected part of the paperswhose detail is originally omitted. Most of them are the parts that I and my groupmate presented in front of the other UROP groups during the course of this project. 25

School of Science Department of Mathematics

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: LI Jiangtao / MATH-PMA Course: UROP2100, Fall

In this report, we will investigate the existence problem of Kähler-Einstein metrics. In section 1, we will introduce some preliminary Kähler geometry and define the first Chern class c1(M) associated to Kähler manifold. In Section 2 of the report, we discuss Shauder estimate for elliptic operators in PDE, which is utilized in later sections. In section 3 we prove the uniqueness and existence of Kähler -Einstein matrics when first Chern class is negative, i.e., c1(M) < 0. The case when first Chern class vanishes, i.e., c1(M) = 0 is similar to the negative case except that the proof of uniform estimate fails. However, with help of Moser iteration, Yau Shing-Tung finally addressed this difficulty. This proof also solves Calabi conjecture, which is known as Calabi-Yau theorem now. The last section will present the proof of Yau, based on the exposition of G.Sźekelyhidi’s “An Introduction to Extremal Kähler Metrics”.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: LIU Lingchong / MATH-PMA Course: UROP1100, Fall UROP2100, Summer

The paper THREE-MANIFOLDS WITH POSITIVE RICCI CURVATURE written by Richard S.Hamiltom used similar idea as in solving curve shortening flow then obtained the conclusion that a compact 3- manifold will have constant positive curvature after Ricci-flow and rescaling. This report is mainly written to fill in details which may be non-trivial for beginners. Also the report could be regard as a kind of notes to summarize the important points and will be helpful for me to recall the content in future.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: MA Yat Long / MATH Course: UROP1000, Summer

The purpose of this report is to review two papers, which are the paper ”The Heat Equation Shrinking Convex Plane Curves” written by M.Gage and H.S.Hamilton, and the paper ”Curve shortening makes convex curves circular” respectively. In order to make the make the paper more comprehendible for general undergraduate students, the report would discuss some results and lemmas in the paper that required more explanation and provide the necessary details, mainly on elaborating the calculations and explaining the equations used in parts 3.1 and 4 of the first paper and the lemmas and theorems stated in the second paper.

26

School of Science Department of Mathematics

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: RUBAB Tamzid Morshed / SSCI Course: UROP1000, Summer

This report adds details to the article by Gage and Hamilton (1986) and the article by Gage (1984) respectively. The former proves that the curve shortening flow on a smooth closed convex plane curve can continue until the curve encloses an area of 0. And the latter proves that under this process, 𝐿2/𝐴 goes to 4π, where L and A are the length and area respectively. This combined with the previous result will imply that the length also converges to 0, so the curve becomes a point and under a normalization to keep the enclosed area constant, the curve converges to a circle.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: TAN Kin Aun / MATH-PMA Course: UROP1100, Fall UROP2100, Summer

The aim of this report is to explain an important theorem in Hamilton’s paper: If a compact Riemannian 3-manifold X admits a metric with positive Ricci curvautre, then X also admits a metric of constant positive curvature. If we further assume X is simply-connected, then X is homeomorphic to a 3-sphere, which is a special case of the famous Poincar´e conjecture. Hamilton was the first to introduce the method of Ricci flow and his paper laid the groundwork for the topologists to prove the Geometrization conjecture and ultimately the Poincar´e conjecture. The research was done by studying his paper and this report serves to aid general understanding of the Ricci flow approach.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: TRAN Duc Huy / SSCI Course: UROP1100, Summer

Curve shortening flow is the process of deforming a smooth curve in 2-dimensional Euclidean space by displacing each point with a velocity proportional to the curvature at that point on the direction normal to the curve. In 1984, M. Gage proved that a smooth convex curve evolving under curve shortening flow will converge to a circle when its area shrinks to 0. Later in 1986, he and R. S. Hamilton proved the existence of a solution until the area shrinks to 0, thus completing the proof that it will become circular. In this report, I will outline the structure of the proofs of these theorems, and fill the omitted details in the original papers to make the proofs more accessible to the general readers.

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: XIA Yusen / MATH Course: UROP1100, Summer In this report, I will present the paper THREE MANIFOLDS WITH POSITIVE RICCI CURVATURE written by Richard S. Hamilton in 1982. In this paper, Hamilton introduced the Ricci Flow, which nowadays becomes more and more important in geometric analysis, and stated that a 3-manifold with a positive ricci curvature initially will converge to constant ricci curvature under the evolution of ricci flow. 27

School of Science Department of Mathematics

Geometric Flows Supervisor: FONG Tsz Ho / MATH Student: YUNG Chun Nok / RMBI Course: UROP1100, Fall

Given a simple closed convex curve in ℝ2 evolving through time under the heat equation ∂F /∂t = kN, we would like to show that it remains convex and shrinks to a point (with zero area) eventually by considering its curvature k and deriving an evolution equation. This paper first gives a slightly altered proof from The Heat Equation Shrinking Convex Plan Curves by M. Gage and R. S. Hamilton, with some of the omitted proofs and details recovered. Then, we discuss how a slight generalization in the evolution equation, with some additional assumptions and conditions, still preserves the boundedness of k.

Retrieval of Aerosol Optical Depth using Machine Learning Algorithm Supervisor: FUNG Jimmy Chi Hung / MATH Co-Supervisor: LU Xingcheng / ENVR Student: CAI Shi Zhan / DSCT Course: UROP1100, Summer

In this project, we focused on the Fine Suspended Particle (PM2.5). PM2.5 as a significant air pollution factor will cause respiratory diseases after people breath, resulting in economic loss. Thus, we want to measure PM2.5 accurately. However, we measure PM2.5 relies on the meteorological stations, which are located sparsely. We cannot get the PM2.5 data in the area which has no station. Our goal is to build a model using several variables such as Aerosol Optical Depth (AOD) to predict PM2.5. We made several models and checked the correlation coefficients between predict-PM2.5 and observe-PM2.5. The correlation coefficients are expected to be 0.70 in summer and 0.85 in winter due to the experiment.

Retrieval of Aerosol Optical Depth using Machine Learning Algorithm Supervisor: FUNG Jimmy Chi Hung / MATH Co-Supervisor: LU Xingcheng / ENVR Student: TANG Yutian / DSCT Course: UROP1000, Summer

The poor air quality in China, especially in north China, has always been a public health concern since the early 2000s. In 2013, 'smog' became the keyword of the year. As an important component in smog, PM2.5 can cause adverse health impacts on human beings. AOD, temperature, and other parameters in the Merra-2 dataset are significant indicators for predicting PM2.5. Thus, in this project, I planned to utilize a machine-learning algorithm to build a reconstruction network to encode the Merra-2 data. I built a reconstruction network which has an encoder and a decoder. Because the original data in Merra-2 was huge. It can be extracted after an encoding-decoding process while the reduced information is sent to the network for learning, which can significantly reduce the calculation workload for further training. This reconstruction network can help us match MODIS (Moderate Resolution Imaging Spectroradiometer), Merra, PM2.5, and PM10 then train the model between them in the future. 28

School of Science Department of Mathematics

Use Machine Learning Technique to Retrieve Air Quality Data Supervisor: FUNG Jimmy Chi Hung / MATH Co-Supervisor: LU Xingcheng / ENVR Student: HU Mingyun / MATH-AM Course: UROP1100, Spring UROP2100, Summer

Nitrogen dioxide (NO2) formed from fossil fuel combustion is one of the air's major pollutants. Based on satellite NO2 data and meteorological variables, this study used an ensemble back-propagation neural network (BPNN) to retrieve five-year (Jan.2013-Sept.2017) surface NO2 concentrations over China in 0.25°×0.25° resolution. The 10-fold cross-validation R2 of the ensemble BPNN model was 0.85 based on monthly average data, and the correlation coefficient reached 0.88 between annual average measurements and yearly average predictions. According to annual mean surface NO2 concentration spatial maps, the peak region of NO2 concentration spanned eastern, northern, and northeast of China, while the overall distribution tended to decrease over the years. We also plotted the seasonal NO2 concentration time series for each province of China. The diagram showed that surface NO2 concentrations were highest in winter and lowest in summer.

Use Machine Learning Technique to Retrieve Air Quality Data Supervisor: FUNG Jimmy Chi Hung / MATH Co-Supervisor: LU Xingcheng / ENVR Student: SHA Yu Hin / COSC Course: UROP1000, Summer UROP2100, Fall

Due to the adverse effects on human health and a potential hindrance to development, air pollution has always been one of the most researched topics. Among the field, pollutant concentration forecast has been crucial, since knowing the conditions beforehand allows countermeasures to cope with the detrimental environment. Methods in forecasting could be categorized into numerical and statistical approaches. Due to advancements in computational power, neural networks have been thoroughly studied and applied for this task. In the past projects, an LSTM-MLP based neural network was constructed and tested with a PM2.5 forecast. This project continues the previous works by refining the models and deploys the model for PM2.5 and O3 forecast for 180 stations throughout China.

Use Machine Learning Technique to Retrieve Air Quality Data Supervisor: FUNG Jimmy Chi Hung / MATH Co-Supervisor: LU Xingcheng / ENVR Student: SUN Haochen / DSCT Course: UROP1000, Summer UROP1100, Spring

The project explores applying several deep learning algorithms based on Long Short Term Memory to air quality prediction, namely, using the 24 hours' data, facilitated with the meteorological forecast, to predict other pollutants' concentration for the next 24 hours. Using ground observation data and the WRF-CMAQ forecasting system results, the algorithms can achieve relatively high precision in terms of several metrics (Pearson correlation, L1-loss, and L2- loss). Depending on the distribution abundance of data in different situations, varying complexity models can be applied, such as using 1D-CNN layers before the LSTM layers.

29

School of Science Department of Mathematics Use Machine Learning Technique to Retrieve Air Quality Data Supervisor: FUNG Jimmy Chi Hung / MATH Co-Supervisor: LU Xingcheng / ENVR Student: WU Zhiang / SENG Course: UROP1000, Summer

Nitrogen dioxide (NO2) is one of the most important factors of air pollution. NO2 also participates in other chemical reactions that generate air pollutions, which will further influence the air quality. Therefore, measurement and prediction for NO2 are necessary. Based on our intuition, NO2 has a strong relationship with some meteorological data. Meanwhile, there are lots of statistical methods and non-statistical methods used in capturing these kinds of relations. In the UROP, BP neural network, Support Vector Machine (SVM), and Random Forest (RF) are used. Different combinations of data pre-processing are tried. After finishing all experiments, BP neural network performs better under most circumstances. However, the best correlation coefficient (0.90) occurs in an unexpected situation.

Modelling Neural Circuit Dynamics in Whole-Brain Imaging Data of Larval Zebrafish Supervisor: HU Yu / MATH Student: CHAU Yu Hei / SSCI ZHANG Ke / SSCI Course: UROP1100, Summer UROP1100, Summer

This report is based on our UROP project, which is about the analysis of whole-brain imaging data of larval zebrafish. The aim of this report is to outline a summary of the methodology of analyzing biological processes that we have learnt in the course of this project. By and large, the methods stated below are well-established and common mathematical tools for data analysis. When the information of the zebrafish’s neuronal and physiological behavior has been isolated, further analysis by MATLAB using mathematical methods such as regression and dimensionality reduction will be done. The objective is to isolate or figure out patterns in the neuronal behaviour.

RNN for Biological Neural Networks Supervisor: HU Yu / MATH Student: CHOI Sehyun / COMP Course: UROP1100, Summer

Artificial Recurrent Neural Networks are growing more famous than ever in the recent era due to its large successes in machine learning fields. Its application to biological neural networks has been growing in popularity as well. However, the conventional algorithms might not be a viable solution for training neural networks for biological research purposes. Two alternative algorithms, FORCE and Full FORCE, were reviewed and shown to be plausible methods for training RNN for various tasks by successful experimentation on different tasks for both the biological and machine learning field. On the other hand, a task of reproduction of the results from a recent paper was not entirely a success for its results were inconsistent with that of the original experiment.

30

School of Science Department of Mathematics Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: CHOY Sin Hang Sonia / SSCI Course: UROP1100, Summer

This report first introduces the basics of cluster algebra, before focusing on a paper by Musiker and Schiffler (2009), which gives a formula for the Laurent polynomial expansion of cluster algebras associated to the triangulations of unpunctured surfaces. The expansion is expressed in terms of the number of perfect matchings of the graph of tiles, which is defined recursively through gluing tiles together. In particular, we prove that in the case of triangulations of a cylinder with two marked points, the number of perfect matchings of the corresponding graph of tiles is given by alternating terms of the Fibonacci sequence, which is verified by the corresponding cluster mutation.

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: FOO Peace / MATH Course: UROP1100, Summer

We give an overview of cluster algebras and examine the relationship between some geometric constructions associated with cluster algebras on surfaces. One of these, the tile graph, will be used to derive a direct formula for the Laurent polynomial expansion of cluster variables in these cluster algebras in terms of the perfect matchings of such a graph. An example will be used to link these geometric structures to a well-known recurrence relation in the theory of cluster algebras.

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: FU Wanying / MATH-PMA Course: UROP1100, Fall

The study of Cluster Algebra is motivated by problem of total positivity. This project aims to understand the basics of Cluster Algebra. It starts with several motivating examples to introduce the concept of total positivity. By summarising those properties, the formal definition of a cluster algebra is given, which is later generalised with semifields as coefficients. In the end, it proves the Laurent phenomenon of Cluster Algebra.

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: MAN Ryuichi / MATH-IRE Course: UROP1100, Spring UROP2100, Summer

In this project, we mainly study the structure of 𝒜(𝐷4); the cluster algebra of type D4; from two perspectives. The first perspective is the finite type property of 𝒜(𝐷4). We apply the techniques introduced by Fomin-Zelevinsky (2003) to compute the cluster variables in 𝒜(𝐷4); and further study the symmetry of 𝒜(𝐷4) by studying the exchange graph under folding. In the second part, we follow Speyer-Williams (2005), Scott (2006) and Brodsky-Ceballos-Labbé (2015) to study the relations between 𝒜(𝐷4) and the Grassmannian 𝐺r(3, 6) via the pseudotriangulation model. We show that the combinatorial types of the fan 𝐹3,6 defined by the structure of 𝐺r (3, 6) are preserved under the folding of 𝒜(𝐷4).

31

School of Science Department of Mathematics

Cluster Algebra Supervisor: IP Ivan Chi Ho / MATH Student: SAMIN Thanic Nur / MATH Course: UROP1100, Summer

In this report we discuss Cluster Algebra and a result on cluster expansion formulas on quivers defined on triangulation of surfaces using perfect matching. A cluster algebra is an algebra defined on a set of variables which we call a cluster. The individual variables can be ‘mutated’ through a set of rules, depending on quivers assigned to the variables. These quivers naturally arise from triangulation of surfaces, and we can assign a mutated variable to the paths of the surface. Explicitly, we apply the result to a torus to obtain a variable, and apply mutation techniques to verify the result.

Quantum Groups Supervisor: IP Ivan Chi Ho / MATH Student: AU Kam Cheong / MATH-PMA Course: UROP1100, Spring

Classical dilogarithm 𝐿i2(𝑥) appears frequently in various branch of mathematics. 𝑞𝑞-version of dilogarithm arises recently, and is important in quantum cluster algebra. In this UROP project, I am reading Kashaev-Nakanishi’s 'Classical and quantum dilogarithm identities' (2011), which has a lucid explanation on how cluster algebra produces dilogarithm identities. The main and novel part of the paper is it 'proves' quantum version of dilogarithm identity reduces into the classical case when 𝑞 → 1. However, strictly speaking, the approach does not constitute a proof, but rather an indication that this is very likely to be true. I will address this problem in the last section.

Quantum Groups Supervisor: IP Ivan Chi Ho / MATH Student: PACHECO RODRIGUEZ Wemp Santiago / MATH-IRE Course: UROP1100, Summer

In this report we will study the relation between knot invariants and quantum groups. We will first study the link invariants arising from representations of the Artin braid 𝑛-group via quasi-triangular Hopf algebras, specifically in the case of quantum groups, following the approach done by Turaev (1988). Then, we will study ribbon Hopf algebras and their induced colored ribbon graphs as well as their connection with the category of finite dimensional representations of the underlying algebra, which serves as a foundation for the construction of the invariants of 3-manifold arising from quantum groups developed by Reshetikhin and Turaev (1990, 1991).

Integro-Differential Equations: Theory and Applications Supervisor: JIN Tianling / MATH Student: LI Xuanang / MATH-PMA Course: UROP2100, Fall

In this report, we try to prove the Hopf lemma holds for fractional Laplacian and apply this lemma in moving plane method to prove symmetry and monotonicity of eigenfunction. We also discuss about a few basic properties of the first eigenfunction of fractional Laplacian. 32

School of Science Department of Mathematics Integro-Differential Equations: Theory and Applications Supervisor: JIN Tianling / MATH Student: PACHECO RODRIGUEZ Wemp Santiago / MATH-IRE Course: UROP1100, Fall

In this report, we study maximum principles for anti-symmetric super-solutions of the eigenvalue problem (−Δ)s u = cu in open and bounded domains. Afterward, we employ these maximum principles to study the monotonicity and symmetry of the first eigenfunction for the eigenvalue problem with Dirichlet boundary conditions for the fractional Laplacian in the interval (−1, 1) using the method of moving planes. Finally, we employ this technique to prove that the first eigenfunction is concave in the interval (−1/2, 1/2) by making use of the monotonicity and the maximum principles.

Efficient Algorithms for Visualizing Dynamical Systems Supervisor: LEUNG Shing Yu / MATH Student: QIN Hanmo / MATH-CS Course: UROP1100, Summer In this paper, we will use finite-time Lyapunov Exponent (FTLE) to study the 2D Logistic map, as 2D logistic map is a typical chaotic system, while FTLE being a strong item to analyze the complex phenomenon occurred in chaotic systems. We will observe the process where steps go to infinite and different results of the map with different parameter 𝜆𝜆 when the number of steps gets close to infinity. In other words, we will observe the changing process of FTLE as step number increases, loosely estimate and compare the situations of FTLE when the number of steps is infinite with different 𝜆𝜆, which is the comparison of Lyapunov Exponent (LE) actually. Efficient Algorithms for Visualizing Dynamical Systems Supervisor: LEUNG Shing Yu / MATH Student: WAN Yuxuan / CPEG Course: UROP1100, Summer This report presents an adaptive algorithm to calculate finite-time Lyapunov exponent (FTLE) for the region surrounding LCS. The algorithm is compared with the traditional Lagrangian approach and shows the results of numerical examples based on double-gyre flow. The computational time improves significantly and is demonstrated both in theory and in practice. Further improvements should focus on the choice selection metrics for points passed to the next iteration. A better selection metric should give a more accurate refinement Efficient Numerical Methods for Dynamic Interface Supervisor: LEUNG Shing Yu / MATH Student: TIN Wei Lung / MAE Course: UROP2100, Fall In order to trace how the data points will move in the next second, we need to find the fittest curvature so as to know the norm at those points. In this project, we generate 5 to 8 data points based on small arc of a circle at the beginning. We use spiral to predict the fittest curvature of the data points. Then we generate data points based on parabola to see if we can still find the fittest part of the spiral. Finally, we will put noise into the data points so that we can figure out how random the data points can be.

33

School of Science Department of Mathematics 34 Numerical Methods for Solving PDEs on Surfaces Supervisor: LEUNG Shing Yu / MATH Student: MAK Chun Cheong / MAEC Course: UROP2100, Fall The eikonal equation on surface is used to calculate the traveling time of the geodesic of a particular curved surface. Most of the numerical methods of solving eikonal equation involved manifold, Fast Marching Method and Fast Sweeping Method. In this report, we try to use embedding idea and linearize the usual eikonal equation to solve the eikonal equation on surface. This report introduces a new numerical scheme and further investigate some possible problems behind. Lattice-based Cryptography, Pseudorandom Functions and Algebraic Number Theory Supervisor: XIONG Maosheng / MATH Student: SIT Hiu Chi / MATH-IRE WU Ka Lok / MATH-PMA Course: UROP1000, Summer UROP1000, Summer The Lenstra-Lenstra-Lovász Basis Reduction Algorithm, or commonly known as the LLL algorithm, is a polynomial time algorithm that approximates the set of short vectors in the basis. It is a breakthrough in solving an important class of problem in lattices, the Shortest Vector Problem. In this report, we begin by providing some mathematical background in understanding the algorithm. We then look into the applications of LLL algorithm in solving the NTRU lattices in the Cryptosystem, which is known to be quantum-proof, and in crypanalysis of RSA cryptosystem under relaxed condition. We would also examine some algorithms or variants of algorithms based on the LLL algorithm. Lastly, we would explore some recent advancement in the area, the complexity on the generalization of the LLL algorithm in solving Module Lattices. Some Optimization Problems Supervisor: ZHANG Hai / MATH Student: GURUNG Nishan / MATH-GM Course: UROP1000, Summer This report introduces curve-fitting optimization using Least Squares method with examples of a straight line, higher degree polynomials, and exponential functions as curve-fitting models. In addition, it studies the optimization of Linear SVM in Hard and Soft Margin. After a brief introduction to Clustering, this paper shifts its focus to K-Means Clustering where it provides the clustering algorithm along with a proof of its convergence. Moreover, this paper includes a mathematical derivation for centroid optimization and a heuristic solution for the K-means initialization problem. Furthermore, this report aims to draw a connection between K-means and other optimization algorithms, namely Gradient-descent, EM, and Newton Optimization. Lastly, this report demonstrates a real-world application of clustering in Spam email detection and highlights major differences between supervised and unsupervised learning models. School of Science Department of Mathematics 35 Research in AI and Machine Learning Supervisor: ZHANG Tong / MATH Student: KEH Sedrick Scott So / DSCT Course: UROP1100, Spring In recent years, Transformer architectures have been used very widely in language modelling tasks. However, many such models take up a lot of space, limiting the window range of the self-attention layer. In this project, we aim to look for new ways to overcome this problem and allow for longer range dependencies. As the project is still ongoing, this progress report will simply give an overview of the models we explored, and share some of the preliminary results we achieved while trying to replicate the results of various models. Research in AI and Machine Learning Supervisor: ZHANG Tong / MATH Student: LI Xinyuan / SBM Course: UROP1100, Summer In the summer UROP project, I learned basic knowledge about computer vision through online course CS231n and conducted experiments using Faster r-CNN on pascal voc dataset and visualized the results. Through the course, I grasped basic pipeline and rational for 2D object detection and then read concurrent papers about modern detection methods: fast r-CNN, faster r-CNN, YOLO, mask r-CNN, Focal loss for object detection (RetinaNet) and end-to-end object detection with transformers. Then I chose one of the methods (Faster r-CNN) and implement it through practically writing the codes in Google Colab to build up its functions and visualize the results. Research in AI and Machine Learning Supervisor: ZHANG Tong / MATH Student: TSUI Yuk Hang / SSCI Course: UROP1100, Summer Constructing a Variational Autoencoder (VAE) with high accuracy and without supervision is important for future artificial intelligence development. In the few previous decades, different methods have proposed to reconstruct the model to improve the accuracy of the VAE model. In this review paper, a basic VAE model is constructed to investigate how different settings affect and improve the performance of the model. Literature review is the majority part of the report and different idea of other models will be considered to improve the model in the future. It aims to learn the strength of different models and combine to form a model with higher accuracy. School of Science Department of Ocean Science 36 Department of Ocean Science Study on the Adaptation and Physiological Responses of Synechococcus Isolates from Different Oceans Supervisor: LIU Hongbin / OCES Co-Supervisor: LEE Pui Yin / OCES Student: IO Kai Hong / ENVS Course: UROP1100, Fall Synechococcus is one of the most widely distributed and abundant photosynthetic cyanobacteria. Despite being small in size, it plays an important role in contributing to the global oceanic primary production. Studying of their adaptation to environment changes may help to better understand their distribution, their roles in the biogeochemical cycle and responses to the changing marine environments. Species in the genus Synechococcus are genetically and phenotypical diverse. Four strains of Synechococcus isolates from different oceans were studied. The continue absorption spectra of each strains under white light and red light adaptation were measured. Chromatic adaptation responses to different wavelength of lights were focused in this study. Study on the Adaptation and Physiological Responses of Synechococcus Isolates from Different Oceans Supervisor: LIU Hongbin / OCES Co-Supervisor: LEE Pui Yin / OCES Student: IO Kai Hong / ENVS LUK Hau Ching / BISC Course: UROP2100, Spring UROP3100, Summer UROP1100, Spring UROP2100, Summer Synechococcus is one of the most widely distributed and abundant photosynthetic cyanobacteria. Despite being small in size, it plays an important role in contributing to the global oceanic primary production. Studying of their adaptation to environment changes may help to better understand their distribution, their roles in the biogeochemical cycle and responses to the changing marine environments. Species in the genus Synechococcus are genetically and phenotypically diverse. Fourteen strains of Synechococcus isolates from different oceans were studied. The continue absorption spectrum of each strain under white light adaptation, grown in media with different Nitrogen source. Each strain is attempted to be grown in media with nitrate ion, ammonium ion and both nitrate and ammonium ion respectively. School of Science Department of Ocean Science 37 Water Quality Survey for Hong Kong's Marine Fish Farming Zone Supervisor: LIU Hongbin / OCES Co-Supervisor: LAU Wing Keung / OCES Student: FU Jiayi / EVMT Course: UROP1100, Spring Hong Kong government has been conducting marine water quality inspection since the mid-1980s. The Hong Kong government needs to monitor the water quantity especially in Marine Fish Farming Zone to maintain the qualified properties for fisheries; thereby, protect the good market condition and people's livelihood. Hong Kong University of Science and Technology, together with Agriculture, Fisheries, and Conservation Department, conducts the monthly survey of the water quality of the marine fish farming locations around Hong Kong's coastal waters. Hydrographic, water chemistry, and biological parameters, including temperature, salinity, dissolved oxygen, total suspended solids, chlorophyll concentrations, inorganic nitrogen, E.coli, and bacteria abundance are measured. This report focuses on the monthly monitoring result in four different zones, including Bluff Island, Round Island, West Lamma Channel, Sha Chau. We would focus on the behavior of each component in four sites and the reason for the showing pattern including location difference and seasonal effect. Water Quality Survey for Hong Kong's Marine Fish Farming Zone Supervisor: LIU Hongbin / OCES Co-Supervisor: LAU Wing Keung / OCES Student: LEE Hong Yui / BISC Course: UROP1000, Summer The quality of seawater in Hong Kong is constantly monitored, examined and improved, ensuring the local fishery and agricultural industries could operate with qualified conditions, produced high quality products to their customers globally. The Hong Kong Agriculture, Fisheries and Conservation Department (AFCD) cooperated with the Hong Kong University of Science and Technology (HKUST) to inspect and examine the water quality, ensuring both safe products could be produced while the local marine ecosystem could remain the least disturbed. The focuses of this report will be on the amount of suspended solid, salinity, dissolved oxygen (DO), chlorophyll-a concentration and N-nitrite (N-NO2), N-nitrate (N-NO3), N-ammonia (N-NH4), Si-silicate (Si-SiO2) and P-Orthophosphate (P-PO4) according to the data measured from January 2019 to December 2019. The findings and interpretation focus on numerous parameters at four different sites in Hong Kong, included West Lamma Channel (WLC), Bluff Island (BI), Round Island (RI) and Sha Chau (SC) by comparing spital and seasonal differences between one another. School of Science Department of Ocean Science 38 Water Quality Survey for Hong Kong's Marine Fish Farming Zone Supervisor: LIU Hongbin / OCES Co-Supervisor: LAU Wing Keung / OCES Student: LI Yung Ho Coco / BIOT Course: UROP1000, Summer HKUST has been cooperating with the Agriculture, Fisheries and Conservation Department (AFCD) in monitoring the water quality of different marine fish farming zones around Hong Kong’s coastal waters for many years. Water samples were collected from designated fish farming zones regularly. In-situ measurements were done together with sample collection. Parameters such as water temperature, salinity, dissolved oxygen, biochemical oxygen demand (BOD5), pH, and turbidity were measured using a YSI EXO II multi-probe sensor. Water depth was measured by Secchi disk. Analyses of suspended solids (SS), chlorophyll-a (Chl-a), and nutrients (nitrate-N (NO2-N), nitrite-N (NO3-N), ammonia- N (NH3-N), silicate-Si (SiO2-Si), and orthophosphate-P (PO4-P)) were done in the laboratory. Together, these parameters reflect whether the water quality is safe for mariculture. Chaotic Mixing Supervisor: MAK Julian / OCES Student: KANG Zheng Tien / PHYS Course: UROP1000, Summer The project aims to build up a chaotic mixing model which allows us to simulate the spreading of heat in chaotic systems. The heat equation was developed in the meantime. The numerical model is solved within the spectral framework, utilizing the Fast-Fourier Transform (FFT) algorithm to compute derivatives. The backward-Euler and Adams-Bashforth two-step multistep method to solve the heat equation numerically. We began by exploring NumPy, SciPy and Matplotlib, the 3 scientific computing libraries used in the Python language coding process. One result was that the resolution of the FFT frequency spectrum is highly dependent on the condition that the frequency domain is a multiple of the frequency of the constituent wave(s). Chaotic Mixing Supervisor: MAK Julian / OCES Student: SAEGUSA Haruki / SENG Course: UROP1000, Summer This project looks to utilize numerical models to explain how fluid particles exhibit chaotic behaviour within a dynamical systems context. A numerical approach focusing first on the diffusion aspect was first considered, focusing on time stepping schemes such as Forward Euler (FE), Adams-Bashforth (AB), Backward Euler (BE) and Crank-Nicolson (CN), examined under Julia and in Python for comparison within a spectral method framework. The work forms the basis to include advective effects in the future, as well as consider the mixing of passive and active tracers. School of Science Department of Ocean Science 39 Characterising Coral Reef Function across Anthropogenic Gradients Supervisor: WYATT Alex / OCES Co-Supervisor: PISAPIA Chiara / OCES Student: CHU Yui Kei / SSCI Course: UROP1100, Summer Coral reefs are under the constant threat of cyclones and bleaching events, which might damage corals and disturb net ecosystem calcification (NEC), net ecosystem primary production (NEP). To the impact of these disturbance, NEC, NEP, R and community structure across the reef in Lizard Island were studied in September 2018 and are compared to data from last decade and 2016. The daytime NEC has decreased in 2016, but then recovered in 2018. Similarly, Coral coverage initially declined by 58.3—64.3% but has then recovered. Both NEP and R have decreased compared to 2016.Tropical cyclones and bleaching events would disturb NEC and NEP and threatens coral over the reef and the reef might have grown more resilient to bleaching events. Characterising Coral Reef Function across Anthropogenic Gradients Supervisor: WYATT Alex / OCES Co-Supervisor: PISAPIA Chiara / OCES Student: CHUNG Shun Lap Benedict / ENVS Course: UROP1100, Summer In recent years, the emergence of algal domination and loss in biodiversity over coral reefs are getting more common over various major reef sites (AF. Haanset al, 2014) due to physical disturbance resulted from possibly anthropogenic activities that adversely affected the environment. Turning a vibrant and colorful reef that hosts the foundation of the marine productivity and species richness into a monotonic turf-ful site. Here, a reef site in Lizard Island National Park, at the Great Barrier Reef (GBR) in Australia, has recently experienced bleaching event twice, due to severe warming events and cyclones, was brought into study (Pisapia et al. 2019). A one full day monitoring on Net Ecosystem Calcification (NEC), Net Ecosystem Production (NEP) and a species composition measurement has taken place to measure its functionality with comparison with past figures from other studies. It was observed that the reef system maintained rather the same NEC value and was undergoing the process of recovery in terms of percentage coral cover, however, the primary productivity was low and the main cover was still algal turf over dead corals. Hence it was suspected that the growth of algal turf may have also contributed to hypoxia and climatic events do impose an observable stress on benthic reef ecosystems. School of Science Department of Physics 40 Department of Physics Fourier Optics Supervisor: ALTMAN Michael Scott / PHYS Student: LO Tsz Yu / SSCI Course: UROP1000, Summer This project is about low energy electron microscopy (LEEM) image simulation of one-dimensional objects using the Fourier optics approach. In the project, step phase and amplitude objects and a sinusoidal phase object are simulated. The simulations are performed using a Python-based program. One of the major results is that defocus has a much larger effect on the contrast produced by the step phase object, compared to the step amplitude object. Moreover, overshooting occurs only at the stronger side of a step amplitude object, independent from the sign of the defocus. The contrast produced for step phase objects depends upon the magnitude of the phase and the sign of the defocus. These results help to interpret image contrast observed using LEEM, for instance the contrast produced by atomic steps at surfaces and rippled graphene. Knowing the defocus will have much more effect on phase contrast, changing the defocus during observation can help us to understand whether there is a phase contrast or an amplitude contrast. The Geometry and Physics of Wrinkling Supervisor: ALTMAN Michael Scott / PHYS Student: CUI Zhenhao / SSCI Course: UROP1100, Summer There are two ways to generate wrinkles on a sheet, one is to apply a tensile stress along the length, and another is to apply a compressive stress along the width. For a free-standing graphene flake, these stresses are controlled by the boundary conditions. In this project we aim to investigate the wrinkles of free-standing graphene by using low electron energy microscopy (LEEM), and finally hydrogenate it to graphane. By comparing their wrinkle patterns, we can find how the hydrogenation modifies the elastic properties of graphene. Now we are able image the free-standing graphene over trenches on an insulating SiN substrate with LEEM. And the cleaning of the sample is in progress. We also measured how the surface potential changes when a transverse voltage is applied to the graphene. No-collapse Quantum Mechanics Supervisor: CHEN Tian Wen / PHYS Student: CHAN Siu Chung / MAEC Course: UROP1100, Summer Since the birth of many-worlds interpretation, physicists have attempted to derive the Born rule from the basic principles of quantum mechanics. However, most of the approaches based on relative frequencies have been shown to be unsuccessful. This paper reviews the work of Farhi-Goldstone-Gutmann and argues that it is impossible to replace the probabilistic postulate with the weaker postulate of definite outcome by means of the frequency measurement of an ensemble of infinite copies of a simple system. School of Science Department of Physics 41 No-collapse Quantum Mechanics Supervisor: CHEN Tian Wen / PHYS Student: JHA Shashwat / PHYS Course: UROP1100, Spring The Born Rule in quantum mechanics is taken as a postulate in the “standard” orthodox Copenhagen interpretation of quantum mechanics. We studied an approach by Farhi, Goldstone, and Gutmann, on deriving the Born rule based on the frequency interpretation of probability. This is done by detailing and expanding their work —first a detailed proof of the quantum Weak Frequency Law is presented, and then the process of obtaining the measure from the transformation function is outlined in a particular context. Quantum Optics and Atomic Physics Supervisor: DU Shengwang / PHYS Student: FOO Shi Quan / PHYS Course: UROP1100, Fall UROP2100, Spring The Einstein-Podolsky Rosen paradox (EPR paradox) is a thought experiment in quantum mechanics theory as it violates the special relativity, that is no information can be transmitted at the speed larger than the speed of light. Now, it can be resolved by the no-signalling theorem. This para-dox is related to the quantum entanglement. This quantum entanglement concept is widely applied in various areas, such as quantum teleportation, quantum communication, quantum computations, and improved microscopes. In this report, I would talk about what rules that we need to obey in quantum communication channel and how these features help us to prevent someone steal the information from the channel without exposing himself. Besides, I would also talk about how we use quantum entanglement in quantum teleportation, which is the transfer of a quantum state from a place to another. This task is very hard to achieve without quantum entanglement. This is because any quantum state is impossible to be cloned. This report illustrates the principle behind the quantum teleportation and the Vienna experiment done in 2004 that is about the quantum teleportation is real environment across 800 meters. In that experiment, we have a remarkable result, that is the fidelity is up to 90% for 45◦ polarized photons. Collective Motions and Phase Transitions in Systems Composed of Robotic Supervisor: HAN Yilong / PHYS Student: WONG Pak Hung Philip / PHYS Course: UROP1100, Fall Different from passive Brownian particle, active particle is able to perform active motion, thus active particle system exhibits many interesting behaviors. Here we model active particle with the seeds of grass, which perform active motion when placed on vibration table. We first investigate the origin of the active motion, results show that active motion depends on several parameters, while the surface pattern of seeds is the main cause of active motion. Then we focus on the collective behaviors of the system made up of seeds. By tracking their trajectories on the vibration table, we show that typical boundary collective behaviors and clustering are observed in this model. School of Science Department of Physics 42 Polycrystal-glass Transition in 2D Granular Systems Supervisor: HAN Yilong / PHYS Student: LI Haoyu / PHYS-IRE Course: UROP1100, Summer An experimental setup with linear motors has been built and used to compress a crystal composed of hard and soft granular particles. During the compression-induced crystal-glass transition, particles’ positions have been recorded and measured by the image analysis. The structural quantities, including the radial distribution function, structure factor and orientation correlation function, are calculated from particles’ positions. These results shown a transition point between crystal and glass. In further experiment, this transition point will be better identified, and a larger system will be employed to give more precise results. Building an Apparatus Cooling and Trapping Erbrium Quantum Gas Supervisor: JO Gyu Boong / PHYS Student: YUAN Weijun / PHYS-IRE Course: UROP1100, Fall UROP2100, Spring We are working on building the apparatus to cool and trapping erbium atom. Currently, we are trying to achieve quantum degeneracy. In UROP1100, we have presented some progress in our experiment. Here, the plan of creating a 2D optical trap is discussed as we want to use the 2D optical trap to create the 2D dipolar superfluid. There are different ways to create the 2D dipole trap and we have decided to implement the accordion lattice. The report will review different methods to create the 2D optical trap and compare them with the accordion lattice. Then, the optical scheme is described in the report. Machine Learning aided Detection of the Quantum State in an Atomic Quantum Simulator Supervisor: JO Gyu Boong / PHYS Student: MAK Ting Hin / PHYS-IRE Course: UROP1100, Spring UROP2100, Summer A convolutional neural network (CNN) is applied to identify the topological quantum phase transition in a system of 173Yb atoms trapped in a one-dimensional optical Raman lattice. The results of the previous report suggest that timeof- flight (TOF) may affect the transition point in the low two-photon detuning regime. In this report, the effect of the TOF value on the topological invariant that determines the phase transition is accounted in terms of spin texture. Furthermore, the effect of various preprocessing technique on training the CNN is investigated. The results give insights for training a more effective and predictive model to determine the transition points. School of Science Department of Physics 43 Electromagnetic and Acoustic Metasurfaces Supervisor: LI Jensen Tsan Hang / PHYS Student: AU Wang To Anthony / PHYS-IRE Course: UROP1100, Summer The development of acoustic metamaterials has been accelerating at an unprecedented rate recently, from active metamaterials to non-hermitian structures (Zangeneh-Nejad et al, 2019) to programmable acoustic metamaterials (Cho et al, 2020) and much more. The field of active acoustic metamaterials is of particular interest for this report, and some of the latest developments in this area will be discussed and examined. Our interest lies in the complete modulation of sound waves by means of using metamaterials with an emphasis on resonant frequency control, amplitude modification, frequency conversion and time variance. Due to recent unfavorable circumstances to conduct experiments in person, much of the previously planned tests have been postponed but progress will be resumed as soon as conditions allow for them. The following report will give an account of some work done by others that have inspired future plans, and also the current progress in our research and technical challenges faced. Electromagnetic and Acoustic Metasurfaces Supervisor: LI Jensen Tsan Hang / PHYS Student: KWOK Yuk Lam / PHYS-IRE Course: UROP1000, Summer The natural frequency of torsional mode in cantilever is first investigated. It is calculated and compared with the simulated results. Geometrical properties, including length, height, and depth of the cantilever, is adjusted to reduce the difference between calculated and simulated frequencies. Then, a three-dimensional metamaterial formed by cutting and folding a thin plate is discussed. The function is to produce anisotropic effective mass density, so that the flexural wave propagating in x and y directions may have different performance. More specifically, flexural wave would propagate in one direction but not the other in a certain frequency range. It can be used for sub-wavelength flexural wave manipulation. Electromagnetic and Acoustic Metasurfaces Supervisor: LI Jensen Tsan Hang / PHYS Student: YANG Sandra Sang Xiao / PHYS Course: UROP1000, Summer Light behaves differently as it passes through different materials due to their complex refractive indices. Some light is reflected, some is refracted and transmitted. At each boundary, the reflectance and the phase change of reflected light vary with the material and the wavelength of light. Upon normal incident, these two characteristics depend on two frequency-dependent parameters: the index of refraction n and the extinction coefficient k. This report describes the properties of visible light propagating normally from air to different metals films in terms of their reflectance and their phase angles. The optical properties of brass (Cu-Zn alloy) and electrum (Au-Ag alloy) are compared with that of its composite metals. School of Science Department of Physics 44 Time Varying Metamaterials Supervisor: LI Jensen Tsan Hang / PHYS Student: FAN Wai Hin / PHYS Course: UROP1100, Summer This report aims to describe the construction of a time-varying, digitally virtualized acoustic metamaterial and study its reflective and transmissive properties. We apply amplitude modulation in the time domain to a digitally virtualized passive metamaterial previously realized by Choonlae Cho et. al., examine the response of individual metaatoms and use numerical simulation in the frequency domain to calculate the transmission and reflection coefficients of the overall structure. We note some of the immediately obvious features of the transmission and reflection patterns, including discrete frequency shifts of the input signal and periodicity of the transmission and reflection coefficients with respect to the phase difference between amplitude modulations. Characterize Single-Molecule Electronic Properties using Low-temperature Scanning Tunnelling Microscopy Supervisor: LIN Nian / PHYS Student: LAI Man Wai Katherine / SSCI Course: UROP1100, Summer Low-dimensional structures are of research interest as they exhibit electronic properties that are different than that of bulk materials and may lead to various applications. Simulation is an important tool in understanding the characteristics of different one- and two-dimensional systems. In this paper, we present MATLAB simulation programs written for various systems, such as the one-dimensional chain and two-dimensional square and hexagonal lattices, and we investigate how changes in input parameters affect DOS. Results and trends and features are discussed. We observe that energy levels vary linearly with hopping parameter regardless of system, while DOS vary greatly depending on shape and size. Monte Carlo Simulation of 2D Supramolecular Assembly Supervisor: LIN Nian / PHYS Student: CHENG Man Sze / PHYS Course: UROP1100, Fall Kinetic Monte Carlo (KMC) is a versatile computational algorithm to study the dynamic evolution of surface absorption and diffusion. We use FORTRAN and Python for KMC to generate the result pictures. This report first introduces the foundation of the KMC simulation: how it compares the rate constant to the random numbers and how to set up the simulation parameters. The report likewise discusses the evolution of the atoms on the fcc (111) substrate under different diffusion barriers and bonding energy through 30 sets of data. Last but not least, it also discusses the nonlinear relationships of the number of the nanoclusters (islands) versus the diffusion barriers of the substrate, and the bonding energy between the nearest atoms. School of Science Department of Physics 45 Monte Carlo Simulation of 2D Supramolecular Assembly Supervisor: LIN Nian / PHYS Student: ZHANG Jiachen / PHYS-PM Course: UROP1000, Summer In this project, kinetic Monte Carlo simulation is used to simulate the process of supramolecular assembly on the Au(111) substrate in order to find out the favorable conditions for the atoms and molecules to form connected pattern on the substrate. Several attempts have been achieved through programming in simulation to see how the changing of temperature influence the patterning forming process on the substrate. Programs have also been written to simulate two types of atom and triangle shape molecule pattern formation process on the substrate, but in this report, only the design for the orientation of triangle shape molecules and the bonding rules for them have been settled. Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: CHAN Pak Yin / PHYS Course: UROP1000, Summer In this paper, using the MATLAB program, a system of a two-dimensional supramolecular nano-molecule was simulated. The structure is constructed by the potential pillars which consist of 4 different layers with different potential, each layer contains several circular pillars formed into the shape of a hexagonal lattice. The significant outcome can be calculated, which are the band structure, total density of states (TDOS), and local density of states (LDOS). The goal is to figure out the qualitative relations between the variables by varying the parameters, such as the lattice constant and the pillar potential. The effects of the physical properties of the supramolecular nano-molecule would also explore due to the transformation of the structure. Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: CHAN Wai Ming / PHYS Course: UROP1000, Summer In this paper, a MATLAB program is used to simulate the physical system of a 2-dimensional supramolecular nanomolecule, by numerically solving the 2-dimensional Schrödinger equation. Hexagonal cells with circular potential pillars at center of cells were constructed in MATLAB. From the program, variables such as band structure, total density of states (TDOS) and local density of states (LDOS) can be measured. Then we would observe the variables’ response to the change of parameters, including the physical size and potential of pillars. This paper will also test the complex structure of multi-potential-pillar cell to discover the influence of transformations of structures to the physical property of supramolecular nanomolecules. School of Science Department of Physics 46 Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: CHEN Huijin / PHYS Course: UROP1100, Fall It is shown that when a periodic potential is imposed at a supramolecular surface, surface electron dispersion relation will be modified and an energy gap will appear at the zone boundaries. Scanning tunnelling microscopy can detect this phenomenon by measuring local density of states. To avoid the high cost of the experiments regarding scanning tunnelling microscopy, we use a Matlab program to simulate the electron states. The program uses principles of Blotch’s theorem and numerical methods to solve the Schrodinger’s equation and calculate properties such as energy band, total density of states (TDOS) and local density of states (LDOS). Furthermore, we can change the parameters in the program to discover different relationships between the output properties and input parameters. Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: SEO Jun Young / PHYS Course: UROP1100, Spring Studying the quantum phenomenon of periodic potential imposed at a supramolecule surface using scanning tunnelling microscopes can be inefficient. The MATLAB program can replace scanning tunnelling microscopes to simulate and calculate periodic potential imposed at a supramolecule surface. The program will be used to customize a 2-Dimensional periodic surface with hexagonal cell as its basic cell according to one’s interest and use Bloch's theorem and Schrodinger’s equation to calculate properties such as energy band, TDOS (Total Density of States), and LDOS (Local Density of States). The parameters such as the lattice constant and potential of 2-Dimensional hexagonal cells can be changed accordingly to study the relation between input and output. Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: SHU Zhicheng / PHYS Course: UROP1000, Summer In this paper, physical systems of two-dimensional supramolecular nanomolecules were simulated by numerically solving the two-dimensional Schrödinger equation with certain periodical potentials. Using MATLAB, we construct two-dimensional systems with hexagonal lattices and apply different potentials on the lattices. From the program, the total density of states (TDOS), local density of states (LDOS) and the band structure are calculated. We study the behavior of these variables as we change the dimension and value of the potentials applied. Particularly, one type of potential aroused our interest for having flat band (FB) in its band structure and we studied more about the conditions for observing this phenomenon. School of Science Department of Physics 47 Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: SIU Chi Wang Justin / SSCI Course: UROP1100, Summer In structures with periodic potentials, the electrons are only allowed to have states with certain energies. The electronic density of states (DOS) is a quantity that describes the number of allowed states at each energy level. In this project, the tight-binding model is applied to calculate DOS for different structures. The project aims to inspect the local variation of allowed energy states for various lattice structures. For each atom in the lattice, there are two states with distinct energies. In the model, electrons are allowed to hop from one state to either the other state at the same site or any states of the adjacent sites. Quantum State Calculation of Two-dimensional Supramolecular Nanostructures Supervisor: LIN Nian / PHYS Student: XIA Zihan / COMP Course: UROP1000, Summer In this paper, we simulated several 2-dimensional supramolecular nanostructures using MATLAB, by numerically solving the 2-dimensional Schrödinger equation. This is done first by constructing a unit cell of hexagonal lattice, which consist of circular and triangular pillars in different potential. By solving Schrödinger equation, some significant physical properties can be calculated and shown in graphs, including total density of states (TDOS), band structure, and local density of states (LDOS). Then by varying parameters like lattice constants and potential distribution within a unit hexagonal cell, changes of physical properties can be observed. First-principles Calculations for Real Materials Supervisor: LIU Junwei / PHYS Student: LIU Hanmo / PHYS Course: UROP1100, Fall UROP2100, Spring This semester there are two missions. The first one is building a package for plotting band structure, and the second one is using VASP to calculate the band structure of graphene. The plotting package can produce four kinds of figures: 1. the band structure and its subplots divided by k-points 2. the relative energy distribution of orbitals like s, p, d, etc. and for all atoms of the selected element 3. the relative energy distribution of selected orbitals like s, px, dxy, etc. and for the selected element 4. the DOS of the material The data of the band structure of graphene is gotten through three steps: 1. input the rough structure of the graphene and optimize its structure 2. perform the self-consistent calculation for the optimized structure 3. update the INCAR, POSCAR file and calculate the band structure School of Science Department of Physics 48 Setup of an Automated Telescope System as Sensor Testbench Supervisor: SMOOT George F. / PHYS Student: CHAN Yan Yan / PHYS-IRE Course: UROP1100, Spring This project is to set up an automated telescope system in HKUST for astronomical sensors testing. The first stage of this project is the selection of suitable equipment, which include an optical tube assembly, a Barlow lens, a go-to mount, a camera and a filter wheel. The sky background in HKUST is measured in order to decide the required aperture size of the telescope. This report outlines two approaches to measure the NSB in HKUST and the proposed choices of equipment. Opinion Formation and Voting Models on Social Networks Supervisor: SZETO Kwok Yip / PHYS Student: CHAU Chun Yin / PHYS-IRE Course: UROP1100, Fall We have previously proposed the group H-index as a measure for the influence of a set of nodes on a network. We present further simulation results from exhaustive searches on small networks to support its effectiveness. Motivated a notable result, we hypothesize that the group H-index works well on networks with high H-index assortativity and verify it by simulations. We then show the application of group Hindex on large networks via a genetic algorithm, and demonstrate a few of its benefits over the group degree by examples on some real networks. Charge Generation and Carrier Diffusion in Organic Solar Cells Supervisor: WONG Kam Sing / PHYS Student: LEUNG Yuk Ching / PHYS-IRE Course: UROP1100, Summer The investigation all begins with the discovery of the Stark effect, in which external electric field shifts the absorption of a material. This property eventually implies that the energy level of the sample (HOMO and LUMO) would change depending on the variations on its dipole moment and polarizability. By measuring the variation in transmission under the modulation of alternating electric field, we are able to have a grasp of the Electro-absorbance (EA) graph which is essential to study the electro-optical properties of the materials. The objective of mine is to measure the EA of organic materials with the guidance of my senior, Mister XING Zengshan. The measurements manifested the absorbance and the EA intensity of different wavelengths of light in both 1st and 2nd harmonic of the modulating frequency. In short, the results are consistent and clearly how the EA intensities are changing under different bias voltages according to theoretical predictions. Charge Generation and Carrier Diffusion in Organic Solar Cells Supervisor: WONG Kam Sing / PHYS Student: SEO Jun Young / PHYS Course: UROP1000, Summer With ever increasing demand for clean and renewable energy, scientists around the world are pursuing alternative sources of energy. Organic solar cells are a favorable alternative source of energy because of their low-cost, lightweight and mechanical flexibility. While organic solar cells are available as a viable option, the efficiency of organic cells are lower than their inorganic counterpart. This report will explore the physics and chemistry behind organic solar cells, data collection of two different organic semiconducting materials and what can be done to improve the overall performance of organic solar cells in the future. School of Science Department of Physics 49 Photoexcited Charge Generation and Recombination Dynamics in Perovskite Solar Cells Supervisor: WONG Kam Sing / PHYS Student: WONG Kin To / SSCI Course: UROP1100, Summer Quasi-2D perovskites has attracted more attention with their high environmental stability compared to 3D perovskites. However, most of the works usually focus on the Ruddelsden–Popper (RP) phase q-2D perovskites rather than Dion– Jacobson (DJ) phase q-2D perovskites. In this work, we try to get more understanding of these two types of perovskites by investigating the charge carrier recombination dynamic of RP phase (BA)2MAn-1PbnI3 and DJ phase (BDA)MAn- 1PbnI3 with time-resolved photoluminescence measurement. It is found that the carrier recombination process of both materials exhibits a dependent on excitation intensity but in a opposite manner. The lifetime of (BA)2MAn-1PbnI3 is positively correlated to the excitation intensity because of traps filling while the one involved (BDA)MAn-1PbnI3 is negatively correlated due to the dominant of bimolecular or Auger recombination in the recombination process. Neural Information Processing Supervisor: WONG Michael Kwok Yee / PHYS Student: KWOK Tsz Hong / PHYS Course: UROP3100, Spring This study focuses on the path optimization on the probabilistic landscape, especially on the planet exploration setting. Although there are many path optimization methods, not many of them are suitable for probabilistic landscape optimization. The Modified Elastic Net algorithm, introduced by the previous groupmate, was introduced to be used in probabilistic landscape. The algorithm is implemented to optimize path on a triangular lattice map in this study. It has a good performance generally, that most of the optimal paths achieve a better score comparing with the expected configurations. Dynamic Jamming and Flows of Dense Granular Suspensions Supervisor: XU Qin / PHYS Student: ZHOU Jianhui / PHYS Course: UROP1100, Summer Dense granular suspensions form an important class of complex fluids that exhibits unique non-Newtonian properties. Due to the presence of solid particles, their behaviors deviate from pure liquids under shear stress. In this summer research program, we performed literature reading and simple optical experiments. We utilized a fluorescent microscope (Olympus IX73 Inverted Microscope) to measure the size distribution of zirconium oxide particles, which will be further used in the dense suspension droplets. We have applied the statistical analysis to our results to investigate the granular particles. All the relevant analysis in this project is carried out by using MATLAB program. School of Science Department of Physics 50 Measuring Femtosecond Laser Pulse Duration with FROG Supervisor: ZHANG Jingdi / PHYS Student: LIN Yuchen / SENG Course: UROP1000, Summer The following knowledge and techniques are studied and examined during the summer semester: Non-linear Optics, Non-collinear Optical Parametric Amplifier (NOPA), chirped optical pulse and FROG (Frequency-Resolved Optical Gating). The acquisition of above knowledge serves as a prerequisite for understanding the working mechanism of the FROG and use FROG to do non-trivial measurements. The NOPA and OPA are direct applications of the nonlinear optics and knowledge on Second Harmonic Generation is the critical part in FROG construction, while the chirped pulse can be well characterized using FROG. In the next semester, by using the difference frequency amplification technique with two chirped pulses, a mid-IR pulse with narrow bandwidth and high peak amplitude can be generated. Measuring Femtosecond Laser Pulse Duration with FROG Supervisor: ZHANG Jingdi / PHYS Student: YANG Yuang / PHYS Course: UROP1000, Summer Frequency-resolved Optical Gating (FROG) is the standard way to measure ultrafast laser pulse duration, and it is widely used in labs worldwide to provide convenient and accurate characterization of pulses. In order to successfully implement difference frequency generation of mid-IR pulses and Noncollinear Optical Parametric Amplification (NOPA) for generation of sub-10 fs optical pulses, it is necessary to characterize these pulses in the process of a generation in full. This report reviews a few processes, considerations, and some equations that are of importance when using FROG, including phase matching and dispersion management, and shows numerically simulated Second Harmonic Generation FROG traces of linear chirped pulses (second-order dispersion) and pulses with cubic dispersion. Numerical Simulation of Broadband Terahertz Generation via Laser-induced Air-plasma Supervisor: ZHANG Jingdi / PHYS Student: SO Lap / SSCI Course: UROP1000, Summer This numerical simulation applies Finite-difference time-domain (FDTD) on a 2D plane to simulation broadband terahertz wave generated by a two-color laser- induced air plasma. In the simulation, we first apply a laser intensity profile to drive a local oscillator that in turn, generates the terahertz wave as the incident laser pulse propagates. By applying a varying refractive index and phase angle modulation to the initial phase of radiating local oscillator, the polarity change and carrier-envelope phase (CEP) shift of the terahertz wave is successfully captured. This will hopefully help to obtain insights in developing theory and designing future experiments in relative topics. Abstracts of UROP Projects 2019-20 SCHOOL OF ENGINEERING School of Engineering Department of Chemical & Biological Engineering 52 Department of Chemical & Biological Engineering Development of Functional Food Supervisor: CHAU Ying / CBE Co-supervisor: LIU Yuanshuai / CBE Student: YORDANOV Matey Svetlan / SENG Course: UROP1000, Summer After obtaining highest quality rapana venosa and utilizing latest food processing technology, a functional food capable of enhancing people’s health was developed. Moreover, a new product with similar properties but much cheaper was suggested for improving Global Health. Mimicking Membraneless Organelles with Synthetic Materials Supervisor: CHAU Ying / CBE Student: ALI Shehzore / CEEV Course: UROP1000, Summer Membrane-less organelles (MO) are cellular compartments that dynamically assemble in order to undergo biochemical reactions. Despite the dominant view of intracellular compartmentalization, some early discovered organelles (such as the Nucleolus and Cajal bodies) are not membrane-delimited. The underlying mechanism is Liquidliquid phase separation (LLPS) facilitated by non-covalent assembly of Intrinsically Disordered Proteins (IDP). Under normal physiological conditions, nucleic acids present in the cellular environment have been observed to promote phase separation in many ways. Proteins that contain RNA recognition motif undergo LLPS in the presence of RNA while some can interact synergistically with RNA to lower the saturation concentration needed for LLPS. This report covers the experimental methods used to determine the relationship between RNA concentration and LLPS of polymer-peptide hybrids to establish a tunable means for the mechanism. Designing of the Chemical Process for Polyethylene Furanoate-Based Materials Supervisor: HU Xijun / CBE Student: CHO Subin / CHEM Course: UROP1100, Fall Poly (ethylene 2,5-furandicarboxylate), abbreviated as PEF, has gained the attention owing to its potential to be used as an alternative of poly (ethylene terephthalate), abbreviated as PET. It is a plant-derived bioplastic material, which is recyclable and has superior mechanical, thermal, and barrier properties compared to PET; considering that its properties are often important factors in the production of plastics, PEF would surely become an environmentally friendly alternative of existing packaging materials of todays. However, the conventional polycondensation process of PEF synthesis has several shortcomings involving degradation and discolouration, which derived from the lengthy synthesis process. In an attempt to compensate the undesired outcomes, the project aims to develop an industrialscale production model for the PEF materials by devising a way to reduce PEF production time. School of Engineering Department of Chemical & Biological Engineering 53 Influence of Random Base in Poly(A) tail to the Function of Mammalian mRNA Supervisor: KUANG Becki Yi / CBE Student: LO Shin Dawn / BIEN Course: UROP1100, Fall Messenger RNA (mRNA) delivers genetic information from the to the cytoplasm in protein synthesis. At its 3’ end, mRNA is protected from degradation by a poly(A) tail—a string of bases unimportant to the structure of the protein to be synthesised. Though its name suggests that poly(A) tails consist exclusively of adenine bases, it has been found that other bases may be found in this tail as well. This paper includes a brief literature review on poly(A) tails, experimental procedures involved in mRNA research, and a statistical ANOVA test carried out on results obtained by a fellow researcher. In the test, an attempt was made to investigate the effect of poly(A) tail length on mRNA lifespan. Influence of Random Base in Poly(A) tail to the Function of Mammalian mRNA Supervisor: KUANG Becki Yi / CBE Student: SHI Naiyu / BIEN Course: UROP1100, Spring With the development of sequencing techniques, it has been discovered that some poly(A) tails in human cells are not completely made of adenines. Random insertions of G, U, and C have been found. These insertions seem to have the capability of affecting the half-life and the translation efficiency of the mRNA, however, there is not enough evidence to tell if these insertions are the reason for the changes and if so, how they influence the functions and properties of mRNA. In this project, we will do a literature review to find out the role of the random base insertions, and also suggest future research perspectives to study this topic as well as the possible future applications based on it. Food Quality Assessment by GC-MS and LC-MS Supervisor: LAM Henry Hei Ning / CBE Student: YAP Octavius Jessica / CENG Course: UROP1100, Summer Fish oil is one of the nutrition supplement that is available in the market. It contains a lot of components that are beneficial for health and one of them is omega-3 polyunsaturated fatty acids. Eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) are two of the omega-3 polyunsaturated fatty acids that can be found in fish oil. These components are reported to have lots of health benefits such as preventing some of heart diseases. However, product labeling in the market is sometimes are not genuinely true and might falsely inform consumers. To quantitatively analyzed the real contained of EPA and DHA, free fatty acids contained in the fish oil are transformed into fatty acid methyl esters (FAMEs) and are analyzed by gas chromatography – mass spectrometry. Improving Data Analysis Methods in Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: CHANG Cheuk Yan Atta / BEGBM Course: UROP1000, Summer In tandem mass spectrometry-based proteomics, whether a peptide exists in a biological sample is a key question to be asked. Peptide spectrum matches (PSM) is a solution to peptide identification. The PSMs, however, are not always correct. This project aims to build a machine learning model to separate the correct PSMs from the incorrect PSMs. It generalizes across 16 machine learning algorithms, and compares the accuracy, performance, time taken for the classification. It also finds the feature importance across several models and extract the common features that could possibly be used to build a more efficient model for data analysis on proteomics. School of Engineering Department of Chemical & Biological Engineering 54 Improving Data Analysis Methods in Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: LEUNG Chin Pok / CENG Course: UROP1100, Summer Database searching is used for matching MS spectrums with spectrums of peptide fragments of peptides in proteomics. A score is assigned to each spectrum and the highest score will be the matched peptide. However, the match does not necessarily mean the 2 spectrums are similar. A p value is calculated for the match to estimate the probability of a random match scoring higher than the match. Long has proposed an approach of p value estimation with exponential distributions, an alternative implementation of the method will be discussed. Also, several new features are implemented on the clustering website. Improving Data Analysis Methods in Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: LI He / BEGBM Course: UROP1100, Spring Proteomics is the study of proteins and their functions at specific times and locations in the cell. Mass spectrometry (MS) is a tool for study the peptide sequences by application in de novo sequencing of peptides. Existing do novo sequencing software, PepNovo, can execute the result in a short time but with high fault rate. To improve the method of de novo sequencing using PepNovo, various modifications can be treated to the sample as well as the algorithm. Furthermore, new model may be introduced for higher accuracy. Investigation of Bacterial Antibiotic Persistence by Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: KUMAR Ishita / CBME Course: UROP1100, Summer Several relapsing infections are linked to bacterial persisters - a phenotype where a minority of the cells enter a dormant state and survive antibiotic treatments to then proliferate once the treatment has stopped. To study the resuscitation process, E. coli wild type and ΔelaB mutants cells were treated with ampicillin to isolate persisters. Samples of the persisters were allowed to resuscitate. Using shotgun proteomics, proteins expressed in the samples were identified. Although an insufficient amount of proteins were detected in the wild type cells, the investigation continued to compare the ΔelaB mutants cells. Fifteen differentially expressed proteins were found (7 up-regulated, 8 down-regulated). Studying the proteins showed that ribosomal proteins, chemotaxis, and aerobic respiration are important for persister resuscitation. Investigation of Bacterial Antibiotic Persistence by Proteomics Supervisor: LAM Henry Hei Ning / CBE Student: NGAI Mark Alexander / BEGBM Course: UROP1000, Summer Bacterial persistence is the culprit of different types of chronic infectious diseases that affects millions of people worldwide. Using genomics and proteomics approach to identify and analyze the key players of persisters, the two genes, mdoH and icd in Escherichia coli (E. coli), have been found to be vital for persisters and are summarized in this literature review. The mutation within mdoH gene has been found to increase persistence by promoting the formation of biofilms and also altering the osmoregulation ability of E. coli. In contrast, the mutation in icd was the result of the integration event of a prophage which restored the function of the icd gene via an SOS response after it was disrupted by genetic damage. School of Engineering Department of Chemical & Biological Engineering 55 A Bifunctional Reactor Design for Biological and Chemical Oxidation Processes of the Greywater Remediation Supervisor: LAM Leung Yuk Frank / CBE Student: BALAJI Kavya Geetha / SENG Course: UROP1000, Summer Greywater is defined as the wastewater collected from baths, kitchen sinks, showers and washing machines, excluding that produced by toilets. Greywater constitutes to 50-80% of the total household wastewater. Therefore, recycling and reusing of greywater is seen as a sustainable solution in a generation where water is a scarce resource. There are several physical, chemical and biological constituents in greywater that much be treated before it is deemed fit for consumption. This report provides a detailed explanation into the treatment of greywater using ozonation, which provides the efficiency of removing about 95% of chemical oxygen demand and 96% of E.Coli present in the wastewater. Two different methods for the reduction of nutrients such as nitrogen up to 98% and phosphorus up to 99% are also compared. A Bifunctional Reactor Design for Biological and Chemical Oxidation Processes of the Greywater Remediation Supervisor: LAM Leung Yuk Frank / CBE Student: CHO Subin / CHEM Course: UROP1100, Spring Water is essential for life. The increasing need for guaranteeing an effective and sustainable resource of water has led many governments and regulating officials to make investments in the conservation of the depleting water resources. The treatment and reuse of greywater is often regarded as a key method. The objective of this ongoing project is to implement a grey water purification system, which treats wastewater collected from sinks and baths into clean water with the recyclable qualities and properties. The system would be controlled in an utterly autonomous manner. The computer software will calculate the amounts of chemicals added into the system and add them while handling the operation of the hardware of the system. In this particular project, the total organic carbon (TOC) and phosphorus content of the wastewater sample will be the main focus. The presence of organic carbon and phosphorus in laundry wastewater will be monitored and the removal methods of multiple organic molecules will be explored. A Bifunctional Reactor Design for Biological and Chemical Oxidation Processes of the Greywater Remediation Supervisor: LAM Leung Yuk Frank / CBE Student: TANAKA Cindy Aiko Filbert / CEEV Course: UROP1100, Fall UROP2100, Spring UROP2100, Summer This ongoing project aims to implement a portable greywater purification device that treats greywater collected from laundry machines in student dormitories for further recycling options. In this part of the project, the topic will mainly be focused on the removal of total organic carbon (TOC) and phosphorus content in laundry wastewater. Greywater properties, the presence of carbon and phosphorus in laundry wastewater, multiple TOC and phosphorus removal methods and required chemicals/materials and their prices will be investigated to design an optimum wastewater purification device. A treatment unit consisting of a primary clarifier (sedimentation) tank, biological reactor tank, and ozone contact tower has been designed and experiments are being conducted to optimize the operational parameters to achieve a maximum efficiency in the treatment of laundry greywater. Later, a prototype of the device, combined with the mobile application, will be developed to demonstrate the feasibility of the whole system. School of Engineering Department of Chemical & Biological Engineering 56 A Bifunctional Reactor Design for Biological and Chemical Oxidation Processes of the Greywater Remediation Supervisor: LAM Leung Yuk Frank / CBE Student: TSANG Cho Sin / CEEV Course: UROP1000, Summer The greywater treatment is not very common in Hong Kong. Greywater is wastewater produced from sinks, showers, baths, washing machines or dishwashers etc. The grey water contains soap, soda, detergents and other chemicals for removal of stains, oil, grease and dirt. From some studies, they showed that laundry wastewater has high SS, COD, BOD. After series of water treatment, the laundry wastewater can be recycled for other applications instead of disposal as waste. The characterization of laundry wastewater and treatability studies on the primary wastewater treatment which is the coagulation-flocculation to laundry wastewater will be shown in the report, to help the whole treatment model design, treating laundry wastewater from the school hall. Development of Valuable Fine Chemical Products from Wooden Waste by Heterogeneous Catalytic Pyrolysis Supervisor: LAM Leung Yuk Frank / CBE Co-supervisor: HU Xijun / CBE Student: LIEM Albert / SENG Course: UROP1000, Summer As a metropolitan and one of the most densely populated places in the world, land prices in Hong Kong is exorbitantly high. With limited land supply and increasing demand for space, comprehensive and strategic planning is required to provide much needed capacity for waste disposal in Hong Kong. With the rapid growth in the logistics and construction sectors, those factors have led to a large demand for wood pallets for easy transportation. Typically, pine pallets are preferred due to its low cost and it cheap enough that pine pallets are often only used once before being abandoned at the end of the supply chain, consequently ending up in landfills and contributing to wood waste that occupies 8- 10% of the landfill capacity in Hong Kong. Thus, this project aims to develop specially designed metallic-based catalyst to transform wooden waste to more useful organic compounds such as pyroligneous acid and bio-oil. Extraction and Purification of Stevia Rebaudiana Glycosides (甜菊糖) for the Application as Sweetener Supervisor: LIU Yuanshuai / CBE Student: CHEN Duoli / CENG Course: UROP1100, Fall Stevia rebaudiana glycosides are extracted from Stevia rebaudiana and have been approved to apply natural low calories sweeteners as for food. Rebaudioside D is a minor component of stevia rebaudiana glycosides and is reported to have better taste attributes. The project proposes a method to obtain rebaudioside D in high concentration from the product of preliminary extraction and purification. The method utilizes the slight difference in solubility against temperature with other components to increase the concentration of rebaudioside D in controllable steps. The method also includes suspension steps and recycling streams to increase the yield. School of Engineering Department of Chemical & Biological Engineering 57 Development of Active Graphene Based Blood Purification Material Supervisor: LUO Zhengtang / CBE Student: TSANG Sin Hang / CHEM Course: UROP1000, Summer Cytokines are a board of proteins produced by leukocytes to regulate the immune response to inflammation and infection. However, in some rare cases, sepsis and acute organ failure for instance, are resulted due to overproduction of cytokines, leading to over-activation of the immunity, entitled hyper-cytokinemia. To alleviate such conditions, removing these cytokines from the blood plasma could be an urgent approach. The objective of this research is to produce graphene oxide-chitosan composite and assess its effectiveness in removing cytokines from blood. This kind of composite is chosen bases on its large surface area / high porosity to enhance the accessibility for small molecules, and its biocompatibility to blood. Most importantly, it exhibits high adsorption capacity the cytokines due to the charge nature. Advanced Catalysts for Water Splitting and Hydrogen Production Supervisor: SHAO Minhua / CBE Student: SEO Jun Young / PHYS Course: UROP1100, Fall As pollution caused by conventional energy sources worsens the environment and with the depletable nature of it, there is an ever growing demand for clean and renewable source of energy. Many sources of renewable and clean energy are available today, one of them being hydrogen fuel cells. Hydrogen can be extracted in many ways, either from hydrocarbons or through electrolysis of water but since hydrocarbons are depletable by nature, electrolysis of water is a more promising option. The surplus energy produced by renewable energy sources are not storable, but the energy can be used to power electrolysis and the hydrogen can be stored for later usage. Today, the most widely used electrocatalyst is platinum, but due to its high cost, we will be exploring the alternative of platinum: palladium. The challenges faced by palladium as an electrocatalyst is its relatively lower activity and stability compared to platinum, but modifications can be done on palladium surface to improve on chemical and physical inferiority of palladium as an electrocatalyst. Developing Genetically Encoded Molecular Tools for Controlled Intracellular Phase Separation Supervisor: SUN Fei / CBE Student: NGUYEN Ha Chau / BIOT Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer Bioimaging is a crucial part of research, diagnostics, an in clinical settings. Although there have been developments in non-invasive bioimaging, most of the widely used and available methods consist of ionizing radiation. Furthermore, these methods often lack contrast agents that exhibit any target specificity. Contrast photoacoustics is a novel bioimaging technique that can offer rapid and specific imaging of biological tissues. However, majority of the contrast agents are based on metal nanoparticles or uses low wavelength light to illuminate the contrast agents, limiting the imaging to a shallow depth of field. This project aims to use protein engineering to develop genetically encoded molecular tools for rapid, deep, specific, and biocompatible photoacoustic imaging in complicated biological systems. School of Engineering Department of Chemical & Biological Engineering 58 Development of a Deep Learning Approach for Transferring Gray-Scale Images Acquired from Photoacoustic Imaging to Histological-Stained Images Supervisor: WONG Tsz Wai / CBE Student: KOBASHI Minato / BIEN Course: UROP1000, Summer Convolutional neural network can achieve image to image translation from one style to another, an example is virtual histological training, providing a new approach for histological analysis to obtain stained tissue sample images without laborious staining practices. Deep learning technology opened up countless opportunities in many fields including medical and biological imaging techniques. In this report, we have trained a cycle consistent generative adversarial network (denoted as cycleGAN) by integrating various deep learning techniques to translate gray scale tissue images to a virtually histologically stained tissue image from unpaired image to image training We will also discuss its limitations and methods to improve the model’s accuracy of the generated images. School of Engineering Department of Civil & Environmental Engineering 59 Department of Civil & Environmental Engineering Air-water Flows in Urban Drainage Systems Supervisor: CHAN Shu Ning / CIVL Student: CHIU Chun Hoi / CIVL Course: UROP3100, Spring In drainage and sewerage systems, vertical dropshafts are commonly used to convey water flow from high to low elevations. When the underground tunnel is run surcharged, the lower part of the dropshaft is submerged in water. The impact of falling water on the free surface results in the entrainment of a vast amount of air, leading to the transport or entrapment of air in the tunnel, causing undesirable effects such as the reduction in flow conveyance capacity or air “blow-back”. The mechanisms leading to these air-water interactions remains elusive, primarily due to the lack of detailed observations and measurement. This study aims at understanding the air transport and air entrainment processes in a submerged dropshaft, using a series of physical model experiments. Based on high-speed imaging and pressure measurement, the flow characteristics have been observed to be dependent on the dimensionless dropshaft discharge Q’ = Q/(gDr5)1/2, where Q is the water discharge, Dr is the dropshaft diameter, and g is the gravitational acceleration. The flow can be classified into three regimes: (1) bubbly column flow when Q’ < 0.16; (2) quasi-steady rising slug flow when 0.16 < Q’ < 0.30, and (3) instability and surging flow when Q’ > 0.30. From Fourier transform and autocorrelation analysis of pressure signals measured along the dropshaft, dominant frequencies between 0.13 Hz to 0.25 Hz were observed, indicating that the surging period is about 4-8 s. The work elucidates the flow behaviour under a range of discharge and shaft diameters, and enhances the understanding of air entrainment and flow structures in a submerged dropshaft. (A video showing the flow regime examples in a model dropshaft of 0.05m diameter under different discharges can be found in https://youtu.be/-8insnQRrG0) Regional Impacts of Global Warming on Wind Energy Resources Supervisor: IM Eun Soon / CIVL Student: HA Subin / EVMT Course: UROP1100, Fall Despite the great advantage of wind energy and the rapid growth of the wind power industry, the actual wind power generation is still quite limited due to the uncertainty of wind power predictions. Hence, this project aims to estimate the changes in wind energy potentials under climate change in Jeju Island, South Korea. The ensemble climate projections will be comprehensively analyzed by using the latest climate models. This progress report includes the introduction of the Regional Climate Model system (RegCM), methodology, and my learning progress in preparation for the future tasks related to how to handle a huge amount of climate data and how to analyze and interpret the future climate impacts. School of Engineering Department of Civil & Environmental Engineering 60 Engineering for Safer Urban Forestry Supervisor: LEUNG Anthony / CIVL Student: LEE Lok Wai / CIVL LO Yu Shun / CIVL Course: UROP1000, Summer UROP1000, Summer Tree stability under wind plays an important role to maintaining a sustainable and resilient urban forest. The purpose of this study is to survey the topologies and morphologies of a given tree species to facilitate future modelling of windtree- soil interaction. The growth conditions and the topology of 7 trees of the same species named Ilex rotunda, which three of them were recently transplanted in various places within the campus, were surveyed using the standard tree risk survey form from the Tree Management Office (TMO) and manual measurements. Two out of nine existing trees were found leaning, and nine out of 12 trees have cracks. In general, the newly transplanted trees have more complex topology and morphology than the existing trees. Although the survey form developed by the TMO includes most of the parameters for tree assessment, to perform the risk assessment more objectively, some parameters involving wind conditions may be added. It is also suggested to improve the tree risk assessment form by including more classes of branches so as to obtain a clearer picture of the trees’ structural form. Water Retention of Hydrophobic Soil Supervisor: LEUNG Anthony / CIVL Student: FOK Sum Yin / CIGBM Course: UROP1000, Summer Hydrophilic soil can be rendered hydrophobic by natural processes or artificial treatments. The use of these water repellent soil to reduce water infiltration in geotechnical structures such as slopes and embankments has been proposed by engineers; however, its hydraulic properties are not well understood. Researchers attempted to use the axis-translation technique to obtain its soil water retention curve (SWRC) but the testing technique would force a concave meniscus formed within soil pore throat rather than a convex meniscus expected in hydrophobic soil. In theory, if a convex meniscus exists in hydrophobic soil, an SWRC may be measured under positive pore-water pressure and atmospheric air pressure. The objective of this project is to develop and test a modified pressure plate extractor for the measurement of SWRC of hydrophobic soil under a positive pore-water pressure condition. Fundamentals of Hydrophobic Soil Supervisor: LEUNG Anthony / CIVL Student: LI Sau Long / SENG Course: UROP1000, Summer Hydrophobic soil, which could be utilised to minimise water infiltration, is investigated for its possible use in slope stabilisation. However, how the artificial application of the hydrophobic coating may affect the mechanical behaviour of hydrophobic soil has rarely been studied. The purpose of this study is to detail the apparatus development of a suction-monitored direct shear apparatus, which is aimed to determine the drainage condition of the tests and any effects of suction on the mechanical behaviour of hydrophobic soil. The hydrophobicity in sands was attained by chemically treated with different mass ratios of dimethyldichlorosilane (DMDCS) in order to create a layer of silicone coating, called polydimethylsiloxane (PDMS). In this report, a direct shear apparatus was modified for mounting a tensiometer, so that any change in pore water pressure developed in the polymer-coated soil of varying water content or degree of saturation can be measured and studied. School of Engineering Department of Civil & Environmental Engineering 61 Non-Contact Crack Quantification of SHCCs without Pre-Treatment of the Surface Supervisor: LEUNG Christopher Kin Ying / CIVL Student: GANI Bryan Suryaraso / MAE Course: UROP1100, Spring Concrete is an essential part of modern civil engineering. However, concrete is not invincible to damage that resulted in cracks. The field of computer vision is advantageous in detecting, quantifying, measuring, and possibly preventing the occurrence of cracks on concrete in the future. In this research, Fibre Reinforced Concrete (FRC) of various fibre types and composition were made and analysed. Concrete samples are tested under direct tension to determine the mix composition that can produce the most number of cracks and to see how mix composition affects the ultimate strength. The images of each concrete sample are then segmented and augmented to become dataset for computer vision purposes. Coupling Sulfur-Driven Denitrification with Anammox for Energy-Efficient Wastewater Treatment Supervisor: WU Di / CIVL Co-supervisor: CHEN Guanghao / CIVL Student: CHUI Wai Yuen Ivan / SENG Course: UROP1100, Summer This study aimed to develop a sulfur-driven denitrification and anammox (SMOX) integrated bioprocess for the treatment of Chemically Enhanced Primary Treatment (CEPT) effluent with low environmental impacts. To investigate this new bioprocess, we have set up a lab-scale system, consisting four 1-Litre Moving Bed Biofilm Reactor (MBBR) and feeding with the synthetic wastewater (45 mgTOC/L and 65 mgN/L) at flow rate of 35 L/day, and continuously observed the operational performance over long term. Finally, the findings show: SMOX was successfully developed, with the average organic carbon and total nitrogen pollutants removal rate at 0.7 kgTOC/m3/day and 0.9 kgN/m3/day through sulfate reduction, partial (de)nitrification and anammox reactions. School of Engineering Department of Computer Science & Engineering 62 Department of Computer Science & Engineering AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: CHAN Long Hei / GBUS Course: UROP1000, Summer This paper presents the independent research the author has conducted on indoor object localization. The main focus of this research is the application of smartphone sensors combined with building-specific information such as floor plans to obtain an accurate estimation of the object position. This approach could be helpful for locating indoor cars in parking spaces and pedestrians in common indoor places such as schools and shopping malls. To test this approach, sensor data from different activity scenarios are collected for pattern recognition. Mapping of the data to indoor locations and application of statistical models to generalize the localization algorithm is still in progress. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: CHIU Ka Ho / COSC Course: UROP1100, Fall Indoor localization using Wi-Fi fingerprinting is becoming more and more popular in recent years. ¬Using the detected vectors of RSSIs from different Wi-Fi APs and relative positioning, localization can be achieved with no information required for distance and angles, leading to a better performance. This report is for one of the possible final steps of such technique. Wi-Fi fingerprinting have two phases, survey and query. This report would assume that the first phase, the survey phase is done and some labelled locations with exact coordinates and an RSSI vector is obtained. The focus would be on the second phase, trying to get a model for predict the target position by using similarities between the target position with some known, labelled positions. Since coordinates and RSSIs varies for each labelled positions and target positions, an embedding procedure is need for storing information of each points in lower dimension graphs. As a final step of Wi-Fi fingerprinting method, stress majorization techniques are used for generating predicted, usually two-dimensional coordinates for all target positions based on the obtained graph information after embedding. In this report, two algorithms would be considered and experiment on: Scaling by Majorizing a Complicated Function (SMACOF) and Vivaldi, for Vivaldi, both the simple one and the one considering remote error will be considered. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: GUPTA Abhishek / COMP Course: UROP1100, Summer Object detection is a well studied problem in computer vision which aims to detect relevant objects in an image/video frame and classify them into one of the ground truth classes. Models for object detection can be used to detect people in a video stream. These detections can then be used to assert whether social distancing measures are being followed in the area covered by the video, like the permitted maximum number of people in a room. The aim of this UROP project is to create an accurate system to detect and track people from an input low-resolution CCTV video stream and to incorporate features to analyse the behaviour of the people in relation to the social distancing guidelines. School of Engineering Department of Computer Science & Engineering 63 AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: JIANG Tianrui / COMP Course: UROP1100, Fall In this semester, my task is to continue the project of smart campus, with recording new sets of data and those for constructing the neural network for rough classification of positions. This was planned on November, but the data to be recorded becomes untypical. Another task is to do data mining on the data fetched by metro systems. Currently, it is far from a mature field. The types of information mined from them is very limited, while the tools to do estimation and prediction are upgraded rapidly. In this semester, we try to exploit the data fetched from Shenzhen metro system in the whole year 2016. Long-short Term Memory (LSTM) and Graph Convolutional Networks (GCN) are utilized in this spatial-temporal data prediction problem. Because of a suspension in November and December, the pace was slowed down for a month, but we still kept contacting and updating codes and ideas on Github (https://github.com/zshhans/Spatial-Temporal-Data-Prediction). As the work was not finished, we will continue to about February for a conference to be held then. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: MAK Yuet Hang / COMP Course: UROP2100, Fall This project is about collecting Wifi-fingerprint information via a mobile application. In an attempt to optimise the positioning of the mobile device, GPS signal would be collected and stored in the application and mapped to the internal coordination of the floor planthrough Least-Square method. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: PENG Zhuoxuan / COSC Course: UROP1100, Summer License plate recognition in video streams is a useful application of computer vision with deep learning methods. Our project takes one step further in that panoramic pictures are generated by “stitching” images captured with different cameras, and Raspberry Pi is used as the server for video streaming of the panoramic images. In this project, I am mainly in charge of the implementation of the video encoder and video request handler, two essential components in the process of video streaming. After finishing these, I will attempt to optimize the encoding process with deep learning methods. School of Engineering Department of Computer Science & Engineering 64 AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: SIM Kyu Doun / COMP Course: UROP1100, Spring The following progress summarizes what I have personally learned from a video super resolution project, “AI meets Big Data”. Firstly, I explain a couple of terminologies related to super resolution, and then introduce a naïve way to perform a single image super resolution. Then I would introduce data sets and libraries specifically used for single image super resolution. For the progress of this project, I was able to understand a few popular models and briefly explain those. Later the report is concluded with further improvements and works left. The main concern is how would one be able to efficiently construct a model that could take in video inputs and produce output in a time efficient manner, and how would it be able to create a dataset that could be suitable for testing the performance of the model. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: SZE Kai Tik / COGMB Course: UROP1100, Spring This research aims to demonstrate how can Core Location Service gathers data by using Wi-Fi, GPS and iBeacon in iOSdevice. In this research, basic iOSprogramming is used to demonstrate how to get user’s location and turn it into coordinate, as well as researching paper online to find more principles and ways about Core Location uses. Although the research is still in a very early stage, it could provide a further research possibility to apply Core Location into indoor positioning service. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: TANG Yubo / COMP YIN Yue / COMP Course: UROP1100, Fall UROP2100, Spring UROP2100, Fall This UROP is detecting user physical movement based on the data collected from sensors on one’s mobile phone and the related networks. First, we armed with the essential knowledge about mobile sensors (to be precise, accelerometers and gyroscope sensors). Then I developed an Android application with my partner that collects acceleration information of the mobile phone and analyzes it as required. Then under the instruction of the assistant He, I improved the accuracy of the detection and produced the output of a formatted user movement information of each experiment. Sequentially, provided with a training model, I managed to understand the code and observed the production of the machine learning algorithms with my own input arguments. School of Engineering Department of Computer Science & Engineering 65 AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: XIANG Letian / COMP Course: UROP1100, Spring As crowdsourcing data becomes a new trend, better data-processing architectures are required for greater performance. To make improvements on the existing architectures, especially the Lambda architecture, techniques like Apache Kafka and Apache beam are needed for higher efficiency. By employing these two techniques, we can improve the data feed as well as flatten the lambda architecture, leading to new data-processing architectures with fault tolerance as well as better performance. With such improvements, data can come in and be processed at real time, resulting in higher accuracy and a lower cost. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: ZHANG Daofu / COMP Course: UROP1000, Summer During surgery, doctors will put some metal needles inside the patient’s body. And later on, after the surgery, they will take the metal needles out of the patient’s body. Previously, Doctors manually take records of the number of metal needles previously planted and the number of needles later removed to ensure there is no metal needles left in the patient’s body. And nowadays, to ensure a safer and more reliable surgery, it’s better to use a machine to take records of the needles removed from the body. In this project, we tried several machine learning based methods to solve the problem. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: ZHANG Xinyu / COMP Course: UROP3100, Fall This report reviews a specific branch under Graph Embeddings research field. With the ubiquitous network structure associate with Internet and mobile devices, the research on graph data mining surges and specific attention is paid on the representation problem in Graph. Transforming the interconnected nodes and edges into numeric form, while keeping the network’s connection information and sometimes the semantic meaning of nodes or edge, remains a challenge over the past decades. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: ZHAO Jiachen / COMP Course: UROP1100, Spring UROP2100, Summer My UROP object is to write a program that can track multiple targets in one camera and multiple cameras. The program should calculate the number of people and staying time of each person and set off alarm when people are against social distancing. School of Engineering Department of Computer Science & Engineering 66 AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: ZHOU Yucheng / DSCT Course: UROP2100, Fall The technical advances in the field of localization have made it possible for the collection and storage of large location data from end-users and entities conditioned on their consents as well as certain legal terms. Outdoor positioning can be achieved by Global Positioning System (GPS), and for the indoor counterpart, Wireless-Fidelity (Wi-Fi) is recently adopted because of its more stable performance. Mining these legally acquired location data can be theoretically and practically beneficial for researchers and citizens. This semester, I have worked on two related projects, one of which is about trajectory completion and the other one is bus arrival time prediction. This report introduces their backgrounds, related works, as well as the adopted and proposed methods and implementations. AI meets Big Data: User Analytics and Personalized Recommendation Based on Location Data Supervisor: CHAN Gary Shueng Han / CSE Student: ZHUANG Xiang / COMP Course: UROP1000, Summer In this project I worked with Peter on a state-of-the-art iOS application for COVID-19 contact tracing with the technology of WIFI signal collection. Under the global pandemic over the virus, contact tracing acts as an important role in slowing down the spread of COVID-19. Countries and regions over the world has implemented their own way of contact tracing with different technologies. Considering the current situation that the virus is breaking out in Hong Kong with an average of over 100 people infected per day, the implementation of such an app is especially necessary. Despite the urgent need for an efficient contact tracing app, factors such as privacy is taken into consideration as well as people’s identity or location information can be taken as a risk if the application requires storing such information and handles it incorrectly. The application my team is developing can avoid such exposure of personal information and give a stable and efficient approach to informing potential exposure to the infected. For my work, I did research on the existing technologies for contact tracing, listing their benefits and limitation, and also helped develop the local data storage. Deep Learning for Dyslexia Pre-screening Supervisor: CHAN Gary Shueng Han / CSE Student: TANG Tianhao / COSC YANG Hanwei / RMBI Course: UROP2100, Summer UROP1100, Summer Many children in this world have dyslexia problems, but it is hard to detect it from a medical perspective. Manual tests on dyslexia also cost a lot of time and work. Our project will focus on this problem and try to give a concise and precise way to detect the dyslexia problems of children through a series of games on character recognition using artificial intelligence as assistance. In this report, we will first introduce our project which is committed to developing an application to test the dyslexia problem of children using both machine learning and deep learning method, and then we will introduce our work on each method in detail. School of Engineering Department of Computer Science & Engineering 67 Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: LEUNG Ko Tsun / DSCT Course: UROP1000, Summer A prediction on users’ behaviors based on low-cost MEMS inertial sensors is studied. Aiming at performing indoor localization in MTR stations, various machine learning algorithms are used to predict user activity for the future development of respective speed estimation algorithms on different user activity and subjects. Random Forest Classifier is found to have the highest accuracy after performing experiments and analyzing the result. The result shows that the prediction accuracy can be maintained at higher than 99%. Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: LU Qingyan / COMP Course: UROP2100, Spring Nowadays, there are already some feasible methods to realize ubiquitous localization in a large area, such a big city or district. Wifi fingerprinting is one of the most important mode in those methods. However, there is a problem when implementing this mode. It will consume a lot of time when using the traditional linear search method to find the result in a large area. So, applying some quicker algorithm can improve this mode in a great sense. In this project, we will focus on a space-partitioning data structure to store and search those WIFI fingerprinting data – k-d tree. Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: LU Yijun / COMP Course: UROP1100, Spring We live in the world full of links and connections, and as the connections become larger and more complex in all possible circumstances, people introduce the concept of networks, for instance, social networks. The advantage of networks or graphs includes discrete, straightforward and informative. They usually consist of nodes and links with attributes attached on them. Network embedding, known as a method, to represent every node in a network with low-dimensional vector spaces, is quite popular in many aspects. It generally possesses the characteristics of adaptability, scalability, community aware, low dimensional and continuous. In present, there are lots of techniques developed for network embedding, such as Deepwalk, LINE, Node2vec, which is an extension for Deepwalk and so on. These methods are widely used in areas like visualization, node classification for networks, link prediction and so on, and as for the performance, they provide outstanding capability for relatively smaller networks. In this paper, the technique for larger-scaled information network embedding will be reviewed and discussed, and meanwhile, some evaluation experiments are performed to give people a better understanding of this young technique. School of Engineering Department of Computer Science & Engineering 68 Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: WANG Wenliang / DSCT Course: UROP1100, Fall Although the outdoor localization has already been quite matured over the past few years, indoor localization is still under development. Since GPS (Global Positioning System), the basic development tool for outdoor localization cannot work indoor due to the limitation of satellite access and complicated building structure, indoor localization uses other features such as Wi-Fi signals and magnetic fields to provide localization, tracking and navigation. Compared to outdoor localization, indoor localization has a broader range of applications and an even larger market, so it is of great significance to develop a well-designed indoor localization system. Indoor Localization and Mobile Computing Supervisor: CHAN Gary Shueng Han / CSE Student: ZHU Yuxiang / CPEG Course: UROP2100, Fall The project is about Indoor Localization and Mobile Computing, which is an intense data project. This report will talk about the content management system designed for our indoor localization project, the RESTful APIs to access and modify the content management system and the version control scheme for this system. In this CMS, data is stored as a format called Node, it provides basic version system and “pretty URL” scheme, which we can make use of to provide user-friendly access API and robust version control function. The API schema has been tested and deployed in our project, and the version control function will be published in the next step. Video Analytics and RF People Sensing for Smart City and New Retail Supervisor: CHAN Gary Shueng Han / CSE Student: TANG Tianhao / COMP Course: UROP1100, Fall In the UROP program this time, I learnt skills with respect to locating. More specifically, I studied how to use services provided by Google or use Bluetooth LE with iBeacon to help me do this job. What’s more, this time I learnt some tools and skills associated with Android application development, using these to help to build an interface for both the Google API and the BLE location services. During the program I became more familiar with the Java programming language and some mathematics methods or algorithms. The whole program strengthened my abilities and broadened my horizon. This report will state what I have learnt from the program in details. Decentralized Mobile Applications Supervisor: CHATZOPOULOS Dimitrios / CSE Student: EDRIC Jovan / SENG Course: UROP1000, Summer Centralized systems have been the preferred method for conventional internet applications. The invention of blockchain technology has introduced a decentralized and secure way of managing data. A combination of hashing, encryption and consensus algorithms is involved in the blockchain protocol to ensure a reliable and safe way of storing data. The mechanisms involved in blockchain facilitates a democratic process, whereby consensus between participants determines the management of data, thereby realising the concept of a tamper-proof method of storage. With the advent of blockchain, a new breed of apps has emerged, relying on the security benefits of blockchain in the form of smart contracts. School of Engineering Department of Computer Science & Engineering 69 Decentralized Mobile Applications Supervisor: CHATZOPOULOS Dimitrios / CSE Student: MA Jianheng / SENG Course: UROP1000, Summer Blockchain applications are now rising rapidly in many areas. Because of the special mechanism the blockchain implements, it makes these applications very secure and efficient to use. Meanwhile, the simple demo code in this article easily shows the convenience and easiness of building a smart contract and the feasibility of applying smart contracts in different areas. Practical ML-based Mobile Applications Supervisor: CHATZOPOULOS Dimitrios / CSE Student: RASTOGI Tanya / COMP Course: UROP1100, Summer Every person is dependent on mobile applications for basic services ranging from online banking to ordering food or medicines, where most of these use machine learning. It employs artificial intelligence to improvise user experience by learning from previous data experience. However, we not only use technology to our advantage, but also twist reality and make a cherry-picked version of it. One such adulteration of reality is the deep fake technology. This technology enables the morphing of an individual’s facial features and voice to realistically portray them saying cohesive sentences or enacting actions that the owner of the face had not done. Moreover, the software tool utilizes various facial recognition models and video editing techniques such as convoluted neural networks and feature recognition using python. This paper combines these two major advances and explains the various machine learning models used to create and detect the deep fake media. Additionally, it also provides a flutter based mobile application to detect if a given image is a deep fake which is implemented using android studios. Deep Video Super-Resolution Supervisor: CHEN Qifeng / CSE Student: CHENG Tim Ta-ying / COMP Course: UROP1100, Fall DeepFake videos are video footages where the face of the person in the image has been forged or swapped out with the face of another individual. The aim of this project is to create a classifier that can achieve high accuracy in determining a video has been manipulated. We first did a paper evaluation on DeepFake models, then we moved on to search upon past methods in DeepFake Detection. We then reimplemented a model structure of a CNN with a bidirectional RNN from past researches and swapped out the original Densenet Convolutional network to a EfficientNet-B0. The model was tested with ImageNet, and we are currently resolving the issues regarding the storage of our model on DeepFake videos. School of Engineering Department of Computer Science & Engineering 70 Deep Video Super-Resolution Supervisor: CHEN Qifeng / CSE Student: CHIU Yik Hei / RMBI Course: UROP1100, Spring In this report. I provide an overview of six models for image super-resolution based on convolutional neural networks. Namely SRCNN, ESPCNN, SRGAN, ESRGAN, SinGAN and Learn to Zoom. The first four aim to reconstruct the high-resolution image from a digitally downscaled input by training a network to perform the mapping. They tackle the same problem and therefore can be considered as improvements on each other. SinGAN is not a build specifically for super-resolution, but it has the ability to upscale a single image without other training data. Learn to Zoom train the model on optical zoom taken by camera instead of digitally downscale high resolution image as input which produce very good result. I also conduct experiment on using separable coevolution in super resolution architectures. It shows favorable but inconclusive results. Deep Video Super-Resolution Supervisor: CHEN Qifeng / CSE Student: SU Haochen / COMP Course: UROP1100, Spring UROP1100, Summer This summer, I’m still working on the “HDR Enhancement” program with a PhD student Xing Yazhou. After the data collection in last semester, my work of this summer is mainly doing experiments based on the model of yazhou and changlin. In the first stage of experiment, our focus was mainly on the loss function. With the testing on different hyperparameters and different loss functions including L1 loss, cobi loss, yuv + cobi loss, and yuv + cobi + L1 loss, we found the yuv + cobi + L1 loss with yuv weight [0.3, 0.2, 0.2] works relatively well. In the next stage of experiment, we focused on different types of normalization including batch normalization, instance normalization, group normalization, layer normalization, localresponse normalization and weight normalization. Then it turns out that the instance normalization and group normalization have better performance on the luminosity and artifact of the testing case. After these two stages of experiment, the problem of luminosity has been improved to some extent. But it still needs better performance. And there’s still some color problem on both of the training set and testing set. We tried to deepen the network (from 2 down sampling and 2 up sampling to 4 each). Then the training set turns out to be fit quite well. But there’s still overfitting problem on the testing set. Now yazhou and I are trying to do two things. The first is that training our model on PyNET’s model. This may help us know whether our data is enough or not. The second is trying to figure out which part of PyNET’s structure may contribute to the good performance of luminosity. So we may add that part into our model. Deep Video Super-Resolution Supervisor: CHEN Qifeng / CSE Student: SUN Yushi / COMP Course: UROP1100, Spring This report introduces my learning process on computer vision in the 2020 spring semester and most of its content would be about my learning outcomes and thoughts about existing GAN networks. The report will first explain my thought on the intuition behind the design of GAN network, then it will demonstrate some GAN networks that I have tried during the semester and finally a summary will be given for my UROP study this semester. In general, this report serves as a milestone for my exploration in computer vision and hope that I can make more discovery and progress in this field. School of Engineering Department of Computer Science & Engineering 71 Deep Video Super-Resolution Supervisor: CHEN Qifeng / CSE Student: XIA Yunlei / COMP Course: UROP1100, Spring The main purpose of this UROP project is to learn some basic knowledge about Computer Vision which is a very important part of today’s development of computer and machine learning. Our Instructor, Professor Qifeng Chen, giving us the idea that this project is going to arouse our interest in Computer Vision and let us have an opportunity to contact some projects and develop our ability to do such kind of research. Change Analysis Supervisor: CHENG Siu Wing / CSE Student: REN Zhengtong / SENG Course: UROP1000, Summer A high-quality edge-weighted graph is useful for various data analysis problems. Daitch, Kelner and Spielman (ICML 2009) proposed such a graph that can be computed by quadratic programming. For the special case of onedimensional data, there exists a faster, O(n^2) dynamic programming algorithm based on an implicit representation of the quality measure function. On random data, an algorithm based on an explicit representation of such function runs faster but its runtime could be exponential in theory. We conducted experiments to construct examples that produce running times exponential to the input size. We also investigated the application of the proximity graph in change analysis. Self-Improving Sorter Supervisor: CHENG Siu Wing / CSE Student: LAM Wai Hin / COMP Course: UROP1100, Fall This report is an empirical study on the self-improving sorting algorithm proposed by Ailon et. al., which sorts an input in O (H+n) expected time where H is the entropy of the input distribution. The experiment results confirm the accuracy of this bound. An alternative implementation using approximated optimal binary search trees was also tested. This implementation reduced the training time from O(n^3) to O(n^2) without significant changes in terms of sorting performance. Two methods of pruning the optimal binary search trees which aims to improve the training space and time efficiency is also examined. Self-Improving Sorter Supervisor: CHENG Siu Wing / CSE Student: WANG Yujie / DSCT Course: UROP1100, Fall The self-improving sorter proposed by Ailon et al. consists of two phases: a relatively long training phase and rapid operation phase. In this study, we have developed an efficient way to further improve this sorter by approximating its training phase to be faster but not sacrificing much performance in the operation phase. It is very necessary to ensure the accuracy of the estimated entropy when we test the performance of this approximated sorter. Thus, we further developed a useful formula to calculate an upper bound for the ‘error’ of the estimated entropy derived from the input data with unknown distributions. Our work will contribute to the better use of this self-improving sorter for huge data in a quicker way. School of Engineering Department of Computer Science & Engineering 72 Building a Blockchain and Smart Contract Application Supervisor: CHEUNG Shing Chi / CSE Student: CHONG Cheuk Hei / COMP Course: UROP1000, Summer UROP1100, Spring Since the growth of blockchain technology, the use of a decentralized network to build decentralized apps (dApps) involves in more businesses. Unlike other technologies, the blockchain history was not significantly long. As a consequence, bugs may be found in some decentralized applications in terms of blockchain confirmation provided in some specific architecture. There may be contradictions between the information stored in the database of applications and the blockchain network, which can lead to fatal errors and mislead the customers with wrong outputs. This research project will concentrate on the code review of certain dApps, and building the benchmark framework to check the actions for identifying possible conflicts during the reorganization of the blockchain. Building a Blockchain and Smart Contract Application Supervisor: CHEUNG Shing Chi / CSE Student: LAU Ying Yee Ava / DSCT Course: UROP1100, Spring Blockchain is a distributed ledger for storing a growing list of records. The records are stored in blocks, and they are linked through storing a cryptographic hash of the previous block. Many advantages of blockchain, such as decentralization and immutability, made smart contract - a computer protocol for the enforcement of a contract, possible. With smart contracts, transactions could be done within seconds without verification by trusted third parties. It is predicted that the usage of blockchain and smart contracts will increase dramatically in the future, and related research is crucial for its implementation. This report will set out to investigate the revert mechanism of smart contracts, i.e. functions that handle or prevent errors, make exceptions to the errors, such as reverting the transaction of smart contracts. Building a Blockchain and Smart Contract Application Supervisor: CHEUNG Shing Chi / CSE Student: LI Daiwei / DSCT Course: UROP1100, Summer Blockchain is a modern information system development framework, featuring distributed data storage, point-topoint transfer, consensus method, and code encryption. The blockchain is a distributed ledger and its network system maintains an orderly data block that keeps growing without a center. Ethereum is a decentralized blockchain stage for 'building relentless applications', whereas Ether is the cryptocurrency utilized on this stage. Programs deployed on the Ethereum Virtual Machine (EVM) are commonly alluded to as 'smart contracts'. The frequently used language for composing smart contracts on Ethereum is Solidity. However, many users of Ethereum have experienced huge financial losses due to crucial security vulnerabilities (Jiang et al., 2018). To detect smart contract for security vulnerabilities, ContractFuzzer is invented. School of Engineering Department of Computer Science & Engineering 73 Building a Blockchain and Smart Contract Application Supervisor: CHEUNG Shing Chi / CSE Student: WANG Minyi / DSCT Course: UROP1000, Summer UROP1100, Fall Before a web-based system going live, issues including the security of the web application and the basic functionality of the site should be addressed by web testing. With my fundamental knowledge of blockchain and DApp acquired in my last UROP project, this research works on web testing on DApp under the guidance of Wuqi Zhang. This report will first introduce the background of the project, state the differences between a DApp and a regular web application, along with the testing dependencies that need to be taken care of. Methods adopted in the implementation will be presented afterwards. And it would be my reflections upon the project as well as my research experience at the end. Detecting Deep Learning Software Defects Supervisor: CHEUNG Shing Chi / CSE Student: DU Wei / COSC Course: UROP1100, Fall Deep learning is an emerging computational paradigm that has demonstrated promising results on mission-critical applications such as auto-driving, disease diagnosis, artificial intelligent communication and other popular fields. However, it is difficult to validate deep learning models. A reason that contributes to the validation difficulties is because bugs can reside in third-party deep learning libraries, e.g., FP16 and Eigen. Thus, various methods have been recently devised to validate different libraries in different deep learning frameworks. However, the models are often interacted with several libraries, which is difficult to be analyzed. In this study, we first collected information about the history of upgraded version and source code of different thirdparty library versions. We then apply the recently proposed validation methodology to detect potential inconsistency or bugs in two major deep learning frameworks, TensorFlow and PyTorch. Detecting Deep Learning Software Defects Supervisor: CHEUNG Shing Chi / CSE Student: LU Yijun / COMP Course: UROP1100, Fall Deep learning is one of the most heated topics nowadays, both in academic and industrial aspects. As deep learning being applied to a large number of areas, the needs for such frameworks come as a consequence. TensorFlow, Keras, Caffe, Pytorch and other frameworks nowadays have been widely used by people, with which you can train your models to perform state-of-the-art deep learning with little effort and background. However, together with the convenience, bugs begin to appear in these frameworks, which can be hard to recognize and discover when using them because it is not the codes written by users that go wrong but the libraries and tools in the software. What kinds of bugs are frequently found in these frameworks? What about their root causes? With these knowledges we can have a clearer mind when dealing with the real-time bugs. Therefore, I collected about 70 bug commits from GitHub about TensorFlow, the most well-known and used frameworks to understand some of the bug types and their coordinate root causes. In this study, I basically have 4 different types of bugs and 6 root causes with examples shown in the form of ‘URL, Symptom, Root Cause (specific), Patch, Commit, Fix Version and Buggy Version’ to make it more straightforward. School of Engineering Department of Computer Science & Engineering 74 Detecting Deep Learning Software Defects Supervisor: CHEUNG Shing Chi / CSE Student: XU Yanbo / COSC Course: UROP1000, Summer Deep Learning has gain huge momentum in recent years and has been wildly used in fields including autonomous cars, financial technologies, facial identity, data analytics and so on. With the support of pytorch, an open source machine learning framework, numerous deep neural network (DNN) models have been developed. Therefore, the evaluation of DNN models is crucial. Among many approaches, one important way of evaluation and test is code coverage, a measurement of the number of lines or blocks of a code executed. This report will illustrate a method of measuring code coverage for pytorch with the help of cmake and gcov, which will later be useful in evaluating DNN models based on Pytorch. Optical Character Recognition (OCR) for Text Data Analysis Supervisor: CHUNG Albert Chi Shing / CSE Student: HARJANTO Alvin / COMP HARTONO Kelvin Leonardo / COMP Course: UROP1100, Spring UROP1100, Spring We currently live in the advent of computing, where we heavily rely on machines and algorithms to generate insights and make decisions, ranging from business analytics in companies to unconscious domestic activities through the use of applications. These algorithms feed on one essential commodity: data. Over the recent years, there have been countless challenges and proposed solutions to gathering more data, as more data equates to higher efficiency. These data most commonly appear in the form of text and numbers, as being intuitive yet relatively simple for computers to process. While there exists a large quantity of sources for data, it is important to note that a huge amount of data aren’t accessible digitally, i.e., they are still in traditional forms such as physical papers, spreadsheets, of which computers can’t directly make sense of. Therefore, Optical Character Recognition (OCR) technology becomes more essential every other day to translate data from physical sources to ready-to-process formats, revolutionizing the conventional methods of manual data input by labor, which costs more and are prone to human error, to ultimately achieve higher speed and accuracy. In this report, we will explore some methods and applications of OCR technologies, from text extraction to reading. Analysis of Convex Hulls and Maximal points of Random Point Sers Supervisor: GOLIN Mordecai Jay / CSE Student: PATUPAT Albert John Lalim / DSCT Course: UROP1100, Fall The maximal points of a set S are those that are not dominated by any other point in S. In a recent paper due to Diaz and Golin, the expected number of maximal points of a set of n random points drawn IID from the convolution of two uniform distributions over two-dimensional Lp balls was analyzed with the motivation of understanding behaviour on perturbed versions of the well-studied uniform distribution. The purpose of this study was to extend the suite of tools designed by these authors to possibly simplify and generalize their case-by-case analyses. Specifically, results show that it suffices to analyze the density function near the upper-right perimeter of the support of the convolution to provide an upper bound. School of Engineering Department of Computer Science & Engineering 75 Constructing AIFV Codes Supervisor: GOLIN Mordecai Jay / CSE Student: ZHU Haoren / MAEC Course: UROP1100, Fall Huffman Coding is optimal Fixed-to-Variable coding if every symbol in the given alphabet can only be encoded by one codeword. By constructing a group of code trees between which the code alternates in almost instantaneous time, a symbol can be encoded by more than one codeword and it improves the Huffman Coding. Current researches utilize dynamic programming to construct the optimal AIFV-m trees. However, neither that the algorithm works well form greater than two nor the geometric meaning of AIFV-m coding is well discussed. This report tries to analyze the geometric meaning of AIFV-m coding and summarize several properties of AIFV-m trees. Dynamic Maintenance of Alphabetic Search Trees Supervisor: GOLIN Mordecai Jay / CSE Student: ZHAO Yizhe / COSC Course: UROP1100, Spring An alphabetic search tree for an ordered probability distribution 𝑝𝑝𝑘𝑘 is a coding tree in which the kth codeword is the prefix code of 𝑝𝑝𝑘𝑘 in the left-to-right order. Alphabetic search trees are essentially binary search trees in which data are kept in the leaves. Maintaining perfectly optimal alphabetic search trees is expensive so the most efficient algorithm is to maintain almost optimal alphabetic search trees. In this project, we use the idea of amortized cost changes which is learned from Splay Trees to maintaining almost optimal dynamically changing alphabetic trees under access and insertion operations. We evaluate the feasibility and superiority of this algorithm with different distribution of data. A De-Polarization System for Social Media Supervisor: HUI Pan / CSE Student: CHAO Chung-chi / COGBM Course: UROP1100, Spring People’s opinions are influenced more and more by the echo chambers built by social media algorithms. This report presents the progress of the study on social media impact on people’s opinion. By creating a survey that mimics the echo chamber that people immense in social media, the responses collected can be an indicator of the influence on people’s perception on different topics. Another part of this report presents the progress on the study of tweets centered on the topic of Covid-19 discussion. By running topic modeling on tweets labeled by various sentiment, groups of topics can be inferred for the tweets. Algorithms and Games in Android Devices Supervisor: HUI Pan / CSE Student: CHIU Ping-chen / CPEG Course: UROP1100, Fall Previously, the Flappy Bird game is controlled by clicking the mouse or press the up or down button to control the movement of the bird. The aim of this research is trying to create a model that could change the input form of Flappy Bird game from clicking to finger pressure. There are three different models in total. The first one is a prototype, the second one applies the use of the Dwell Time confirmation technique, and the third one applies Quick Release skill. We would first introduce the background information of this game. Next, we would talk about the details of the three models and give a conclusion. Finally, we would discuss the future expectation of the game. School of Engineering Department of Computer Science & Engineering 76 Algorithms and Games in Android Devices Supervisor: HUI Pan / CSE Student: HO Yi-hsuan / COGBM IM Man Yi / GBUS Course: UROP2100, Spring UROP1100, Spring Social movements are natural by-products of unpopular policies – people with similar ideals gather and try achieving a mutual ultimate goal. In this project, we focus on the Anti-Extradition Law Amendment Bill (Anti-ELAB) Movement in Hong Kong since mid-2019, aiming at finding the key characteristics of a decentralized protest and how information is transmitted through the social media. We hypothesized that there would be frequent interactions among social media users. In terms of word and image processing, there would also be word choices and images used to deliver certain cues to the general public. Up until the report submission, we were in the progress of data collection and visualization on two main social media, Instagram and Twitter. Algorithms and Games in Android Devices Supervisor: HUI Pan / CSE Student: HO Yi-hsuan / COGBM IM Man Yi / GBUS IP Si Hou / SENG Course: UROP3100, Summer UROP2100, Summer UROP1100, Summer Social movements are natural by-products of unpopular policies – people with similar ideals gather and try achieving a mutual ultimate goal. In this project, we focus on the social protests in the Anti- Extradition Law Amendment Bill (Anti-ELAB) Movement in Hong Kong between mid- 2019 to mid-2020. The goal is to investigate and test whether a so-called decentralized protest stands in terms of the visual elements and network structure to elaborate on how information is transmitted through social media. We hypothesized that certain word choices and images were used to deliver cues as well as there were frequent interactions among social media users. Analysing the three main social media platforms in the movement, namely Instagram, LIHKG and Telegram, we proved our hypothesis using various kinds of algorithms and data visualization. Algorithms and Games in Android Devices Supervisor: HUI Pan / CSE Student: LIANG Haoran / DSCT Course: UROP2100, Fall In this research, I implemented an interactive interface to let the users adjust the latent vector and generate their desired results. Generative image models have the problem of not generating the desired result of the users. So, I have made some bars for the users to change different dimensions in the latent vectors, to adjust different features on the generated result of the GAN model. I have tested on different generative image models, including VAE, StyleGAN, DCGAN, and PGGAN, and used multiple VAE methods on extracting the latent vector to find out which dimension is most decisive in generating specific features. School of Engineering Department of Computer Science & Engineering 77 Algorithms and Games in Android Devices Supervisor: HUI Pan / CSE Student: TIAN Xiangan / COMP Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer This report is a brief summary of my research work related to the influence of network latency to the user experience of augmented reality. There is a simple background description of the study of network latency and augmented reality and my motivation to finish the study. Then it described several methods and tools I have tried for building the augmented reality applications and their evaluations. After that the report states the methodologies for testing the influence on user experiences under different intensity of network latencies. And as a new area for me to study, selfreflection is appended as well to introduce my self-learning experiences during the process. Algorithms and Games in Android Devices Supervisor: HUI Pan / CSE Student: ZHU Zherui / SENG Course: UROP1100, Summer In the summer vacation of 2020, to enrich my extra-curriculum life and have a deeper understanding of the subject of computer science. I take the UROP 1100A project: Algorithms and Games in Android Device. In this project, I learn and help to use HTML to create the research website, and use C++ to solve some geometry problems. After the project, I believe I have a preliminary grasp of HTML and have a deeper understanding of python and C++. In the meantime, the research is about to fully investigate what gesture is most suitable for a certain Android system app interface, and after doing some questionnaires I have already found some rules in this project which makes me have a deeper understanding of the Android system. Augmented Reality on Wearable Devices Supervisor: HUI Pan / CSE Student: CHENG Yiren / COGBM Course: UROP1100, Fall This report presents the UROP research project during the fall semester of 2019, leaded by the SyMLab Director, Prof. Pan HUI in CSE Department of The Hong Kong University of Science and Technology. This UROP research project aims at developing a novel and a truly in-shoe force-assisted and IMU-driven sensor measurement system named Footstep++. In this report, we present the design and implementation of this hands-free interaction and a pilot evaluation follows. School of Engineering Department of Computer Science & Engineering 78 Augmented Reality on Wearable Devices Supervisor: HUI Pan / CSE Student: LIANG Haoran / DSCT NIE Fei / COMP Course: UROP3100, Spring UROP1100, Fall In this research, we focus on implementing a new drone control method by controlling the movement only with a small-sized one-hand hardware through a guideline of direct position control. To assess the effectiveness of our method, we also implemented 5 other methods as comparison. We asked participants to complete multiple tasks to test the performance of each method. After evaluating the results and user feedback, we found IMU outperforms and has achieved a relatively high satisfaction among users. And, we implemented an IMU based one hand control model on drone controlling and found it has optimal user feedback. Augmented Reality Technology for Visually Impaired Supervisor: HUI Pan / CSE Student: TSENG Mu-ruei / COSC Course: UROP1100, Spring Visual impairment is one serious problem that many people faced. Different people may suffer at different degrees and require various levels of medical treatments. Medical treatments, however, are relatively expensive regarding the price and not everyone is able to afford the cost. As technologies become more and more advanced in modern society, it opens up new approaches to conquer this problem. The method that we proposed is to apply augmented reality techniques on wearable devices so that we can largely reduce the costs and also risks for those that are needed. This report will discuss the methodology of implementing this idea. Can We Build a Cloud from the Crowd? Cloud Computing on Smartphones Supervisor: HUI Pan / CSE Student: HUANG Jiaxin / COMP Course: UROP1100, Summer As mobile devices become increasingly powerful and ubiquitous, clouds composed of ambient mobile phones are able to provide unprecedented computational capabilities. This report presents the architecture and implementation of AmbientNet, a distributed system that offloads computationally intensive jobs to enormous ambient mobile phones so as to achieve high runtime performance. To leverage the system, programmers decompose computation jobs into distributable atomic tasks by defining a dependency graph. AmbientNet manages the remaining work including dispatching, profiling as well as exception handling. Evaluations show that AmbientNet adapts to various computation scenarios and provides a speedup almost linear to the number of nodes attached. School of Engineering Department of Computer Science & Engineering 79 Can We Build a Cloud from the Crowd? Cloud Computing on Smartphones Supervisor: HUI Pan / CSE Student: LEE Do Hyun / QFIN Course: UROP1100, Fall Nowadays, mobile devices, such as smartphones and tablets, come up with powerful computing capabilities. Newer devices combine multiple cores that have advanced processing capabilities. We try to form a cloud server using given mobile devices by combining the computing resources of several nearby devices. It is also called as cyber-foraging. However, mobile devices and environments they are running are very diverse. For example, there are many different OSs with different versions for different type of devices. We aim to provide a common platform/system that can help develop a massive, real-life, heterogeneous mobile computation grid. In this progress report, I will state which actions we have done to achieve our goal. Human in the Loop Machine Learning Supervisor: HUI Pan / CSE Student: WONG Wai Ning / COGBM Course: UROP1100, Summer An application tutorial or walkthrough can be useful for a first-time user of a mobile app, or any digital interactive user interface (UI), especially for those with complex functionalities. Yet, not all applications include a tutorial, and not all tutorials are effective in enhancing the user experience. In this project, we collect images of visual hints given in apps tutorials across platforms including Android, iOS, web. By crowdsourcing assessment survey through MTurk, we evaluate the effectiveness of the tutorials by the user engagement (UE) levels, in four aspects: Focused Attention (FA), Perceived Usability (PU), Aesthetic Appeal (AA), Reward Factor (RF). After evaluation of the reliability of the datasets, we can then conclude what are the essentials of an effective application tutorial. The result can be contributed to the development of design tools for tutorials. Human in the Loop Machine Learning Supervisor: HUI Pan / CSE Student: ZHOU Jiaxi / DSCT Course: UROP1100, Spring UROP2100, Summer This report will cover the work I have done in the UROP2100 project, human in the loop machine learning in 2020 summer term. The theme of this UROP session in imitation learning, about which I have learnt relative concepts, theories and implementations. As a result, this report will mainly focus on how I completed the imitation learning tasks and the methods I have applied. Human-Drone Interaction Supervisor: HUI Pan / CSE Student: LIEWCHANPATANA Pitchaya / ECON Course: UROP1100, Summer Since the dawn of the digital age, drones have been playing an integral part of our lives. From commercial to delivery to military purposes, drones have the ability to entertain us with camera recordings, facilitate e-commerce operations and reach terrains unreachable by humans. Yet, existing drone-to-human interactions have fatal flaws. One main issue deals with the drone controller interface- it is difficult to operate the drone. Over the summer, I have worked with Dr. Pan Hui and Dr. Lik-Hang Lee to investigate how one-handed user interfaces can better optimize drone control experiences. My work is split into two key areas- conducting literature reviews and exploring drone simulator softwares for virtual operations. School of Engineering Department of Computer Science & Engineering 80 Interactive Voice Agent Supervisor: HUI Pan / CSE Student: NG Hon Wa / IS Course: UROP1100, Spring The paper discusses the chatbot technology and introduce Meena, a multi-turn open-domain chatbot presented by Google AI. Meena is trained based on data filtered and mined from public domain social media conversations and is a perfect example due to its highly human-like behaviour. In addition, the paper also includes Sensibleness and Specificity Average, which is a human evaluation metric to captures essential component of a human-like conversation. Such metric helps quantitatively analysis and enhance the comparison among different chatbots. In addition to evaluation of chatbot, the paper takes a step further and includes the daily application of voice assistants, which can interpret human speech and respond through text/ synthesized voices. The application shows the implementation of voice assistants and echoed with the new introduced SSA since assistants have to provide relevance and specific responses. Lastly, operation and potential privacy and security Issues which are inevitably will be discussed. Machine Learning on Wearable Devices Supervisor: HUI Pan / CSE Student: SHUM Ka Shun / DSCT Course: UROP1100, Summer Contemporary voice assistants require that objects of interest be specified in spoken commands which is inconvenient for daily usage. Based on the fact that users are often looking directly at the object or place of interest, more information can be used. This report illustrates a software method for smartphones that provides the real-world gaze location of a user that voice agents can respond to command that contains ambiguous words to achieve to a more natural and precise interaction. We achieve this by simultaneously opening the front and rear cameras of an iPhone. The front-facing camera is used to track the gaze pointing in the 2D dimension which is the view of the phone. The rear camera provides the view of the real world. This allows user to intuitively define an object of interest by looking at that region. Since the project is still in progress, there is no user evaluation to test the functional implementation though it can be quick and accurate during the test. Machine Learning on Wearable Devices Supervisor: HUI Pan / CSE Student: WU Chia-wei / CPEG Course: UROP1100, Fall Currently, visual interpretation for convolutional neural network are primarily used as black boxes which do not provide sufficient human explainable reason for prediction results. Interpretation of the module’s decisions are crucial for human end-user to comprehend and justify the credibility of the prediction system. In the work we put forward a human-computer interaction framework called interpretable machine learning. The framework mainly evaluates significance of subspaces lying on images via manipulating Class Activation Mapping method on final convolutional network layer. Afterwards, user can adjust parameters of CNN to optimize prediction result. In our report, we operate our framework on several celebrated neural network, in order to find reliable and interpretable human explanation on neuron networks. School of Engineering Department of Computer Science & Engineering 81 Machine Learning on Wearable Devices Supervisor: HUI Pan / CSE Student: ZHAO Yijie / COMP Course: UROP1100, Fall Mobile headsets enable users to interact with diverse content, in the form of text, images, videos, and 3D objects appearing in an integrated physical and digital environment. Many academic works study interaction with digital overlays of images and videos on the top of the physical world. However, most studies neglect interaction with text content through mobile headsets. On the other hand, the existing commercial approaches for interaction with text content are indirect and ineffective. Users take photos to acquire text information, for example, shown on a poster or advertisement1. Furthermore, there exists no practical way for users to perform direct manipulation on text information in mobile headsets’ web browsers. It is thus necessary to develop systems and techniques for text interaction on mobile headsets. Semantic Distance Learning Supervisor: HUI Pan / CSE Student: LI Yanjia / DSCT Course: UROP1100, Spring From a sociological perspective, polarisation is a social phenomenon that appears when individual align their beliefs in extreme and conflicting positions, with few individuals holding neutral or moderate stands. Thus, as process, it is the increase of such divergence over time when people evaluate issues of diverse nature. This project aims to detect the political polarization and analyse the interaction discourse of users on the social media. Recently, many methodologies have been applied to measure the development of polarization on Twitter, such as opinion propagation model that maps the communication fluxes among the population. In our project we would like to modify the opinion propagation model to see if it could bring better performance in analysing the dynamics of political polarization. First, we build user network based on the hashtags they share with each other. Then, we apply semantic embedding to characterise each user in our social network. We also use semantic distance along with visualisation technique to infer latent attributes of users. We apply this modified methodology to a Twitter conversation about the Indio-Pakistan confrontation and evaluate its performance in detecting different degrees of polarization. Machine Learning Driven Web Log Tracing System Supervisor: IEONG Sze Chung Ricci / CSE Co-supervisor: WANG Tao / CSE Student: CHOI Hoi Sing Viky / COMP Course: UROP2100, Fall The quick advancement of the web has caused companies and organizations to embrace web applications to supply administrations. This has made web application a profitable target for malicious users. In spite of protective instrument such as web application firewall and interruption location framework has been conveyed to fortify the security of web applications, security breach has still been continually observed on these web applications which caused colossal prudent misplaced and undermine the protection of clients included. In this project, we have deployed a model which applies 1-gram analysis for feature extraction and applied a byte distribution pattern comparison to categorize patterns such that through models generated from labeled dataset, we could develop a model which could quickly and accurate categorize new observations. School of Engineering Department of Computer Science & Engineering 82 Privacy Protected Big Data Sharing Environment Supervisor: IEONG Sze Chung Ricci / CSE Co-supervisor: WANG Tao / CSE Student: YAU Tsz Fung / COGBM Course: UROP1100, Fall Analyse the current problems in sharing cybersecurity data in terms of constructing data sharing contract, decision and implementation of anonymizing algorithms. Followed by analysis to big-data log analysis platform, the format of log and data, how logs are being stored and how log formats can be shared between them effectively. Also analyses privacy requirement of log data and determine what should NOT be transferred between log/threats analysis platform, and construct an anonymizing program to implement anonymization algorithm to satisfy the privacy requirement. Deep Coding: Generating Source Code Using Deep Learning Techniques Supervisor: KIM Sung Hun / CSE Student: PRATAMA Kevin / COMP Course: UROP1100, Fall The project proposes an early prototype or implementation of the source code generation using the self-attention Transformer-XL model. The implementation aims to interpret given English natural language and generate a partial source code in Python language based on the intent natural language. The technology stack utilizes the model’s interface in PyTorch, which are fed with input indices corresponding to the natural language tokens, and compared against the output tokens in Python language syntax. By using the model’s interface Transformer-XL, the model is aimed to generate a sequence of snippet token indices. This approach can be treated as a natural language translation approach. Deep Coding: Generating Source Code Using Deep Learning Techniques Supervisor: KIM Sung Hun / CSE Student: ZHOU Xinrui / COGBM Course: UROP1100, Fall This project aims at generating partial source code using deep learning techniques such as RNN and GAN. The professor led weekly meetings where students and research assistants presented what they have learned and implemented for the week. We also together studied some papers in attention mechanism and YOLO in the meetings. The main task for students were studying the materials provided by the professor and research assistants and also implemented and reproduced some paper results or popular deep learning networks. For special reasons, this semester's research project lasted only two months, but we still benefited a lot. Students have different levels of understanding in the field of deep learning, so professors have unique teaching arrangements for everyone. Since I have a background in multiple disciplines and have not been exposed to deep learning, the professor arranged so that I not only have learned a lot of deep learning knowledge, implemented some codes under the help of research assistants, and also made a presentation on the deep learning industry in the project. School of Engineering Department of Computer Science & Engineering 83 Machine Learning Approach to Creating Personalized E-Learning Paths Supervisor: MA Xiaojuan / CSE Student: NIE Fei / COMP Course: UROP1100, Summer One of the most critical skills in our daily communication is to make our arguments persuasive to people. Argument quality plays an important role in enhancing the persuasiveness of the message we are trying to convey. While it is hard for people to develop such skills without feedbacks, in this project, we try to build Persua, a visual interactive system that assists in writing persuasive arguments by providing on-the-fly guidance. In Persua, we evaluate the persuasiveness of the document users' input and provide personalized suggestions on how to modify their arguments by offering convincing example sentences derived from a labelled database based on persuasiveness strategies with more than 1,400 sentences. Users will get visualized feedback on the comparison between their inputs with highquality examples as well as suggested sentences on selected topics. As the project is still in progress, this is a progress UROP report. Machine Learning Approach to Creating Personalized E-Learning Paths Supervisor: MA Xiaojuan / CSE Student: PUYANG Huancheng / COMP Course: UROP1100, Fall Online learning platforms have gained great importance in providing courses on the Internet so that students can learn from anywhere using a computing device on their own preferences. Much flexibility is attached to these processes because of personalization, which enables users to customize their learning paths. Recent years have also seen a rapid proliferation of gamification elements with the expansion of online learning platforms. Effectivity has been shown in using these elements to encourage users to make the most of e-learning tools. To further discover the use of gamification elements, we apply the Artificial Intelligence techniques to explore knowledge on certifications issued by Coursera on LinkedIn accounts, present a course recommendation system for Coursera in a content agnostic way. Machine Learning Approach to Creating Personalized E-Learning Paths Supervisor: MA Xiaojuan / CSE Student: WU Yongjin / COMP Course: UROP1100, Spring Community Question Answering Forum (CQA) is an important tool to acquire knowledge online, and it can be categorized into two basic types by themes – academic and non-academic. One of the critical issues concerning many CQA designers is to understand how user behaviors vary among CQAs of different themes, especially the behaviors of engaged users as they are the main contributors and vital interest of CQAs. To resolve this problem, this project will collect datasets to measure engagement and user behaviors, preliminarily analyze whether the user behaviors in academic and non-academic CQAs is significantly different, and further provide some suggestions for CQA designers. School of Engineering Department of Computer Science & Engineering 84 Machine Learning Approach to Creating Personalized E-Learning Paths Supervisor: MA Xiaojuan / CSE Student: ZHAO Yankun / COMP Course: UROP1100, Spring UROP2100, Summer Active learning is an instructional method in which students actively engage in the learning process instead of being passive recipients of knowledge. In online synchronous classes, active learning helps to involve students in the learning process and achieve better learning outcomes. In this study, we conduct a qualitative research to build a framework tool to characterize student barriers of active learning in the context of online synchronous classes, aiming to benefit the HCI education community, fill the literature gap and help educators to create more inclusive and equitable online active learning classes. Our framework is based on the coding and analysis of data crawled from social networks and responses from semi-structured interviews. Machine Learning Approach to Creating Personalized E-Learning Paths Supervisor: MA Xiaojuan / CSE Student: ZHOU Yuran / CS Course: UROP1100, Summer Community-based Q&A platforms received huge traffic in recent years and were considered by many as the top option to seek for help or suggestions. However, while quality questions stimulate discussions and answers consistently, some other questions were left ignored. Thus, the ability to ask effective questions has become very relevant for the interest of many cQA users. In this report, data from the top three computational and non-computational cQAs will be extracted and studied, some shared traits behind effective questions will be analyzed, the difference in those traits between computational and non-computational cQAs will be discussed, and possible guidelines of asking effective questions in cQAs will be shared eventually. Paper-Based VR/AR Interaction Supervisor: MA Xiaojuan / CSE Student: DING Shanhe / COMP Course: UROP1100, Fall With the rapid growth of technology, researches aiming at enhancing physical paper has experienced a swift rise. Researchers have developed various methods to intensify paper via technological approaches. In our study, we name this kind of enhanced paper the augmented paper. Much as substantial researches about augmented paper have been carried; an all-inclusive summary is lacking, leaving a niche for our project to fit in. Our proposed method essays to provide a comprehensive view of the augmented paper technology. After reading 209 relevant papers thoroughly, we extracted their vital features, including the exploited attributes of the paper, the application domain of the product, the enhanced dimension, and the corresponding implemented technique. School of Engineering Department of Computer Science & Engineering 85 Paper-Based VR/AR Interaction Supervisor: MA Xiaojuan / CSE Student: JIANG Taoxi / COMP Course: UROP1100, Spring Experts are essential for training apprentices, but traditional face-to-face coaching method lacks scalability and is not possible under special conditions. The struggle to deliver realistic coaching experience to mass apprentices has been actively addressed using Augmented Reality technologies. In this study, we summarized some common issues in traditional face-to-face Master-Apprentice training scenarios and some corresponding attempt to rectify these issues using AR technologies. We further extended the comparison between current AR-aided on-site Master-Apprentice training and Remote Master-Apprentice training, where the expert do not have to be in physical proximity with the apprentice. Based on these comparisons, we provide several design recommendations for adopting Master- Apprentice Model for AR Training. Paper-Based VR/AR Interaction Supervisor: MA Xiaojuan / CSE Student: WU Bohuai / COMP Course: UROP2100, Fall Nowadays, there are many different products and research related to augmented paper products. There have been numerous literature papers about the paper augmentation using multiple technology methods. However, it is not clear about how much the augmentation goals can be reached. It is not obvious and easy to measure the functionality of the augmented paper products. Therefore, this report is trying to review nearly 40 different papers relevant to the augmentation for physical paper products. This project is to analyze over 40 paper products and refine the most common and vital attributes that can be achieved by these papers. Finally, we come up with a list of attributes with a list of paper products’ type, which can be used to measure the augmented paper products’ performances. Paper-Based VR/AR Interaction Supervisor: MA Xiaojuan / CSE Student: XIA Zihan / COMP Course: UROP1100, Fall Augmented paper is a term describing paper products which are enhanced by technologies in order to make them more powerful as information carrying medium. Though many works have been published on this topic, a comprehensive understanding of how the capabilities of paper can be extended with technology is currently lacking. Therefore, we present a systematic review of the use of paper and paper products in HCI research with the objectives of mapping out the design space of technologically-augmented paper. We identify the inherent attributes of paper exploited by the HCI community in the design of interactive systems, the existing application domains of augmented paper, the dimensions in which paper can be enhanced with technology, and the implementation approaches used thus far. School of Engineering Department of Computer Science & Engineering 86 Paper-Based VR/AR Interaction Supervisor: MA Xiaojuan / CSE Student: XU Muyu / COMP Course: UROP1100, Spring Communication and collaboration are essential in many daily tasks or in some work scenarios. In recent years, Mixed Reality (MR) has been applied to realize the collaboration that cannot be conducted in real world. With the technologies developing, MR has displayed its potential to enhance the communication between people, and many improvements on the communication techniques have been made. However, the field of communication configuration and spatial scale need more research. In this report, we will discuss both the communication configuration and spatial scale and display some deeper thoughts on them. Real-time Lip Reading through Deep Learning Supervisor: MAK Brian Kan Wing / CSE Student: KINAGI Amrutavarsh Sanganabasappa / COSC KRISHNAMURTHY Padmanabhan / COMP Course: UROP1100, Fall UROP1100, Fall In this progress report, we provide an update on our investigation of novel deep learning architectures and techniques aimed at enabling both visual and audio-visual speech recognition, i.e., audio-independent and audio-supported lip reading. After studying related papers from multiple sources, our group has chosen to modify and implement the paper ‘LARGE-SCALE VISUAL SPEECH RECOGNITION’ owing to its claimed versatility, accuracy and precedent of similarly framed architectures yielding a fair degree of success. The model and its pre-processing pipelines were implemented from scratch, and training has begun on the LRS3-TED dataset. The first section of this report will examine the rationale for choosing this paper, the deviations from the paper present in our implementation, and challenges that are in the process of being overcome. The second section of this report will detail our group’s progress on a parallel task, which is the development of a smartphone application to enable the miniaturisation of the entire visual speech recognition pipeline-right from video capture up to result visualisation. Such a miniaturisation would enable users from outside the deep learning research domain to actually make use of deep learning based lip reading methods, and potentially supplement hearing aids and audio speech recognition. Real-time Lip Reading through Deep Learning Supervisor: MAK Brian Kan Wing / CSE Student: KRISHNAMURTHY Padmanabhan / COMP Course: UROP2100, Spring In this project, I have attempted to investigate, implement and miniaturise deep learning models for visual speech recognition, i.e., lip-reading, in quasi-real-time. Over the course of the project, I have successfully implemented the Visual-to-Phoneme speech recognition model as described in and trained this model on the LRS3 TED dataset. Although the model was trained on a restricted subset of the dataset, I report encouraging phoneme error rates of 34% on the validation set of our restricted dataset, and 68% on the entire LRS3 test set. Parallelly, I have streamlined our lip-reading pipeline to run on a partner iOS App. The iOS app integrates with Amazon Web Services’ EC2 and s3 compute and storage infrastructures to record a speaker’s video, evaluate the model on it, and display predictions, with a latency of only 8 seconds for a 3 second video. School of Engineering Department of Computer Science & Engineering 87 A Study on the Principles and Implementations of Array Databases Supervisor: NG Wilfred Siu Hung / CSE Student: HAU Yiu Tong / COMP Course: UROP1000, Summer Array database management systems (ADBMS) provide an efficient alternative to traditional relational database systems for managing spatial or multidimensional data often seen in scientific applications. As a novel type of database system, it is also important to have a standardized benchmark to evaluate the performance of these systems. This report will introduce the implementation of SciDB, an ADBMS designed for scientific computation, by highlighting its key design features. A detail description of the data processing functions and benchmark queries in Standard Scientific Database Benchmark (SS-DB) will also be given, followed by the preliminary benchmark results for SciDB. A Study on the Principles and Implementations of Array Databases Supervisor: NG Wilfred Siu Hung / CSE Student: HYUN Jeongseok / CPEG Course: UROP1100, Fall In the recent IoT era, tons of data are generated by the machines every second. Multidimensional database has been developed for efficiently managing the machine generated data which are multidimensional data intrinsically. Due to the large scale of data, outsourcing database storage to an external server became a trend and various protocols have been proposed to provide confidentiality of the data and preserve the functionality at the same time. However, such research on multidimensional databases is not yet conducted so that this paper explores the multidimensional database and searchable encryption, then proposes a new multidimensional database equipped with both security and functionality. The experiment is done to measure the performance of range operators on TileDB, Opaque and Spark SQL with the real geographical dataset provided by NCEP Climate Forecast System Reanalysis. A Study on the Principles and Implementations of Array Databases Supervisor: NG Wilfred Siu Hung / CSE Student: ZHANG Weiwen / DSCT Course: UROP1100, Spring The idea of a Multi-dimensional Array Database model has been proposed and implemented in various scientific studies. Different from conventional databases, array, as one of the most ubiquitous elements supporting scientific research in fields like earth science and meteorology, has been considered as the first citizen in ADB. This report mainly covers ideas in Array Database and considers TileDB is an exact instance of Array Database systems to see how Array Database is implemented. Then we briefly study a novel idea, known as Adaptive Repartition, applied in the storage engine of TileDB to increase the performance. Otherwise, some experiments are conducted to present performance of TileDB in the reading, writing and joining performance of TileDB. School of Engineering Department of Computer Science & Engineering 88 An Experimental Study on Incorporating Searchable Encryption into Array Databases Supervisor: NG Wilfred Siu Hung / CSE Student: LIN Ching-yu / COMP Course: UROP1100, Fall A conceptual report for discovering a secured array DBMS in distributed environments. The report consists of three parts, including reviews on the existing variations of Spark, known security issues in the database areas, and the possible methods to implement a secured array DBMS in a distributed environment. The first part mainly discussed current approaches to tuning a well-known distributed in-memory data-base system, Spark, to ingest and manipulate array data, including SciSpark, SparkArray, H5Spark, and SparkSQL. The second part included some reviews on security issues in database management systems. The last part proposed a possible design for a secured ADBMS based on Spark. An Experimental Study on Incorporating Searchable Encryption into Array Databases Supervisor: NG Wilfred Siu Hung / CSE Student: TSUI Ka Kit / COMP Course: UROP1100, Fall Storing and analyzing multidimensional data efficiently, which is inadequately supported by conventional relational DBMSs, is an important problem in past years, resulting in the implementation of many array database management systems (ADBMSs). TileDB is one of the ADBMSs that we interested in the project. The proper order layout of cells in array data and the proper indexing method can greatly affect the performance of read and write operation. However, current designs of physical storage of many ADBMSs, such as TileDB, lack consideration for query profiles, which are potentially pertinent to the overall performance. In this project, we would like to address this deficiency and present adaptive reorganizing of multidimensional bases which caters for tendencies of preceding and ongoing queries. The following will present the basic concept of ADBMS, architecture and feature of TileDB and our proposed new indexing method for TileDB and ADBMSs. Visual Analysis of Big Data Supervisor: QU Huamin / CSE Student: CHEN Xumeng / COMP Course: UROP1100, Fall News articles tend to be increasingly misleading nowadays, preventing readers from making subjective judgments towards certain events. While some machine learning approaches have been proposed to detect misleading news, most of them are black boxes that provide limited help for humans in decision making. In this paper, we present SirenLess, a visual analytical system for misleading news detection by linguistic features. The system features article explorer, a novel interactive tool that integrates news metadata and linguistic features to reveal semantic structures of news articles and facilitate textual analysis. We use SirenLess to analyze 18 news articles from different sources and summarize some helpful patterns for misleading news detection. A user study with journalism professionals and university students is conducted to confirm the usefulness and effectiveness of our system. School of Engineering Department of Computer Science & Engineering 89 Acceleration Techniques for Real-Time Rendering Supervisor: SANDER Pedro / CSE Student: DEMIRER Caner / COMP Course: UROP1100, Summer Increasing standards of resolution and framerate for real-time computer graphics applications bring an increase of workload to GPUs, especially to the fragment shading stage. Re-projection can be used as a means of reducing this workload by caching rendered frames and reusing information from them. When used in its simplest form, it can cause inconsistent frame times due to an expensive refresh of this cache every few frames, thus in real-time applications, it is mostly used in a way that amortizes this cost. This paper describes a new variant that alongside amortizing the refreshing cost, it makes use of the status and properties of each object in the scene in order to increase its efficiency. Efficient Algorithms to Process Gigapixel Images Supervisor: SANDER Pedro / CSE Student: SUGIMOTO Ryusuke / COSC Course: UROP2100, Fall We present a method to generate high-resolution and realistically rendered wave animation from a static image. Our method is separated into three steps. First, we generate an ocean mask. Second, we retrieve parameters and textures used in the third step. We then render the scene with the mask, parameters and textures retrieved in the first two steps. Our approach utilizes cutting-edge deep learning technology for mask and parameter prediction. In the rendering step, the method we developed based on the screen-space reflection method to simulate both local and global reflection from 2D textures is used. For the learning of the texture and parameter prediction network, a differentiable rendering technique is used. Efficient Algorithms to Process Gigapixel Images Supervisor: SANDER Pedro / CSE Student: XU Peiyu / COMP Course: UROP1100, Spring UROP2100, Summer This project presents a way to improve the resolution of an image. Unlike the existing learning based super-resolution models, this project uses additional information from an gigapixel panorama to upsample any image that frames the same scene as the panorama. This session of the project focuses on finding matching points between the highresolution panorama and the low-resolution input image. Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: WU Kaixing / COMP Course: UROP1100, Fall In this project, we are asked to use a Long Short Term Memory (LSTM) neural network to set up a sequence to sequence translator. The purpose of this UROP, according to Hongming Zhang, is to let beginners that are new to research area like me have a taste of research and develop our skills for further research opportunity. During the whole semester, we made use of some Github projects and tutorials to help develop our understanding over neural network, such as CNN, RNN. In spite of the tight schedule, we were still able to get to know the concept of neural network. However, due to the shortened project duration, it is a pity that we were unable to finish the LSTM module. School of Engineering Department of Computer Science & Engineering 90 Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: XIANG Tianqi / COSC Course: UROP1100, Summer Presently reported is a progress report of a research trying to enrich baselines of commonsense reasoning, an area of natural language processing. This project is going to firstly implement this task with Event2mind and ATOMIC that do not have official code. Then it is going to investigate the overlaps between the method of Event2mind and ATOMIC. Finally, it is going to implement COMET as baseline that with official code. This progress report will show the principle of the operations, the procedure of this research, the progress of this project and the further plan. Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: XU Yanbo / COSC Course: UROP1100, Fall Natural Language Processing (NLP) has been found to be an important field in computer science. During the research of NLP, Knowledge Graph (KG) and knowledge base are found to be capable of providing better performance compared with pure text-based models. Among various types of Knowledge Graphs, the Eventuality Knowledge Graph is especially useful. Therefore, automatically completion and generation of an Eventuality Knowledge Graph are of great importance. Furthermore, combined with Computer Vision, Knowledge Graphs could be completed from images and videos, which provides an even more completed knowledge graph. This report will show basic understandings of NLP, KGs and the essential steps in building a sequence-to-sequence NLP model (with KG) and some further applications of it. Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: YAP Zhi Yun / CPGBM Course: UROP1100, Summer Graph Neural Networks (GNNs) have gained rising popularity with its wide application in social network, knowledge graph, and bioinformatics data. With the predefined propagation rules, GNNs are able to capture dependencies in graph-structured input via message passing between the neighboring and local nodes. Previous work on GNNs has been focusing on the node-labeling framework to model intra-cluster similarity and inter-cluster dissimilarity. However, edge information, especially direction, strength, and type of connectivity, is inadequately incorporated. We propose Relational Subgraph Isomorphism Network (RSIN), which is an extension of Relational Graph Isomorphism Network (RGIN) to exploit edge features and capture the rich interaction between entities. New update functions are constructed to enable edge features encoding and inject graph topological information for better graph learning. School of Engineering Department of Computer Science & Engineering 91 Correcting Geometric and Shading Artifacts of Camera Imaged Documents with Deep Learning Supervisor: TSOI Yau Chat / CSE Student: ADIGUNA Joszef Maximillian / COMP LYU Hanfang / MATH-AM SHAO Yuheng / QFIN Course: UROP1100, Spring UROP2100, Summer UROP1100, Spring UROP2100, Summer UROP1100, Spring UROP2100, Summer Document digitization efforts have been made on document rectification tasks. Traditional methods for document digitization focused mainly on correcting distortions using additional information other than the original image. Improvements in the field have allowed the use of deep learning approaches in favor of traditional algorithms, as they have been proven to be more accurate, resulting in higher quality digital images more suitable to digital analysis. We propose that these methods can also be adapted to perform correction on non-text documents, mainly paintings and pictures. Preliminary training using paintings has shown that existing methods work to some extent, correcting certain geometric distortions akin to paintings, albeit with lesser degree of success. Correcting Geometric and Shading Artifacts of Camera Imaged Documents with Deep Learning Supervisor: TSOI Yau Chat / CSE Student: CAO Qinglan / DSCT JIANG Xudong / COSC LU Xinbo / COSC Course: UROP1000, Summer UROP1000, Summer UROP1100, Summer Correcting geometric and shading artifacts of camera imaged documents has been a trending research topic in computer vision, with methods ranging from geometry-based solutions to recent deep learning approaches. In this project, we reviewed some previous research papers on document unwarping and explored their algorithms. We attempt to apply DewarpNet to generalize the idea of document correction from text files to paintings. To this end, we first render painting based distorted document images and other ground truth files as our dataset. Then we train and infer using our data in the DewarpNet with some variations. We obtained some decent results in our experiments, although the effect of correction still is still in need of improvements. Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: HSU Chia-hong / COMP Course: UROP1000, Summer As social media has become people’s major news source, research on detecting fake news has also grown into a viral topic these days. In this paper, we investigate the performance of state of- the-art Graph Neural Networks (GNN) for network-based fake news detection. Besides, we propose an RWR-GAT model that combines the Random Walk with Restart (RWR) mechanism into the training of Graph Attention Network (GAT). RWR was first introduced to take account of the global graph topology, since the over-smoothing phenomenon of GNN is inevitable when more layers are trained. Prior works such as PPNP and PPRGo sequentially design an RWR propagation stage after the training features are finalized by a Graph Convolution Network (GCN) stage. Regarding our RWR-GAT model, we propose an algorithm that trains the global topology feature by RWR during the GAT stage. To the best of our knowledge, no existing algorithms for fake news detection have incorporated the global network structure during training. We then define those previous GNN methods as baselines to evaluate the performance of our RWR-GAT model over the FakeNewsNet dataset. School of Engineering Department of Computer Science & Engineering 92 Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: PATUPAT Albert John Lalim / DSCT Course: UROP1000, Summer Personalized PageRank (PPR) - also known as Random Walk with Restart - is one of the most deeply studied node proximity measures in the graph data mining community. Specifically, for frequently-evolving graphs, it is typically required to re-compute PPR values which immediately reflect changes, and hence, lightweight index-free approximate algorithms are preferred for this use case. The state-of-the-art approach FORA achieves both efficiency and theoretical guarantees, but its theoretical analysis and implementation details show room for improvement. Thus, this study analyzed FORA using a greedy perspective in order to minimize running time. Results show that both parameters of FORA can be improved and several implementation details can be optimized. For future work, this study could motivate faster FORA-like algorithms for computing PPR. Efficient Queries over Database Supervisor: WONG Raymond Chi Wing / CSE Student: SHUM Ka Chun / COSC Course: UROP1100, Fall In the past few decades, 3D technology such as object modeling and visualization has facilitated a great number of useful and valuable applications that could be implemented in the real-world scenes. Researchers and producers are keen on many newly sprouted things like object capturing, object reconstructions, augmented reality and virtual reality. 3D data and its related techniques are widely used in both academic studies and industry. Thanks to the decreasing price of depth sensors, 3D techniques are now easier to be implemented to real-life scenarios. In order to study our HKUST campus deeply, we decided to construct a 3D model of our campus with the help of some state-of-the-art 3D technologies. The construction results are expected to serve as an open-source database, in which users are able to catch the digital information of the campus easily. It would be a great support for data analysis and future applicationrelated developments, which would make our HKUST life more diverse and convenient. Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: ANG Clyde Wesley Si / DSCT Course: UROP1100, Fall Online shopping is a rising industry, and along with it is automated recommendation. In particular, session-based recommendation systems is useful in systems where users are only logged in for a short period of time and we need to predict the user’s preference. In order to better predict the next click of a user, this paper explores the effectiveness of recent preference representation in neural network models. We explore different recent preference models and find a similarly improved result compared to models without recent preference, indicating that recent preference representation in session-based neural network recommendation systems can indeed improve prediction results. School of Engineering Department of Computer Science & Engineering 93 Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: CHENG Tien-chun / COMP WANG Weiqi / COSC YU Yiduo / COMP ZOU Yiwen / COMP Course: UROP1100, Fall UROP2100, Fall UROP2100, Fall UROP1100, Fall In this project, we aim to create an online version of terrain toolkit. The webpage accepts a 3D geographical file as input. We can visualize it on our canvas and convert it into a triangulated geodesic graph. The 3.js library is widely used in the JavaScript part to draw the 3D model and make it interactive. The webpage also supports two functions, simplifying the graph and computing the shortest path. In our implementation of the shortest path algorithm, we apply the CGAL library, which is due to Xin and Wang's approach. The overall interface is programmed by HTML5 and CSS with the help of JavaScript. We have partial progress in implementation. Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: DO Thuy Trang / DSCT Course: UROP1100, Summer Voronoi diagram is a popular data structure that has many applications nowadays in different fields especially in solving some queries based on geo-textual data where each object with some keywords is located in a certain space. Voronoi diagram can be considered as an approach to interpolate the spatial data into many polygons such that each point in a certain polygon is closer to the generator point of that polygon than to the generator point of others. This report is devoted to the construction of Voronoi diagram and its application in the paper 'On Nearby-Fit Spatial Keyword Queries'. This allows us to have comprehensive background knowledge about Voronoi diagram, so we could understand the important contribution of this data structure to solve the problem in the paper. Knowledge Discovery over Database Supervisor: WONG Raymond Chi Wing / CSE Student: ZENG Junyan / COMP Course: UROP1100, Fall The task of session-based recommendation is using a dataset with no user ID to compute the rank of probability of the next item. This model is needed in online shopping website to help services better to customers. There are some different methods to achieve the goal. The better model that makes better prediction of customers’ choices, will get the interest of customers easily and persuade them to spend more. So, solving this problem not only has technical value but also can make profit directly which shows the reason why this topic is so hot these days. School of Engineering Department of Computer Science & Engineering 94 Research on Mining Course Structure Supervisor: WONG Raymond Chi Wing / CSE Student: HSU Shang-ling / COSC Course: UROP2100, Summer Predicting grades with some student’s and/or course’s information is a crucial task in educational data mining. Much research has been conducted to solve the problem, and some of them are based on linear regression, matrix factorization, other empirical models, sequence models, attentive models, and graph models. Nonetheless, not only does these methods have their respective problems, but they also share some issues, including sequential state updates, shared information, co-taken courses, and explainability. Fortunately, these limitations can be analogized to some similar situations in the sequential recommendation task, so the method used to address the problems may also be adapted. In the meanwhile, addressing such limitations may bring advance to the sequential recommendation task, too. Research on Mining Course Structure Supervisor: WONG Raymond Chi Wing / CSE Student: LIU Ziming / SENG Course: UROP1000, Summer In this UROP 1000 project, students should perform data mining tasks on all important components of courses on a given list to extract their structure and correlation among different courses. I am instructed to study the basic knowledge: C++, data structure and basic algorithms, to be able to read and understand the research papers, and thus lay the foundation for future research in the field of data mining and database. Gesture Recognition Using Raspberry Pi Platform Leveraging Machine Learning Technology Supervisor: ZHANG Qian / CSE Student: CHOU Chih-an / COMP Course: UROP1100, Spring Gesture recognition has been widely used for various applications in the world. For example, people could change TV channels by making hand gestures, so that the direct contact to a remote control or the television become unnecessary, or gesture recognition technology could be used as recognizing sign languages, which may improve the communication with deaf-mute people. This study was conducted to determine the hand gesture by using OpenCV and mathematical tools. While other studies related to detecting hand gestures utilize finger and palm segmentation like, this study detected hand gestures by using contours of hands, convex hull, convexity defects, and law of cosine to recognize the number of fingers showing in front of the camera. School of Engineering Department of Computer Science & Engineering 95 Gesture Recognition Using Raspberry Pi Platform Leveraging Machine Learning Technology Supervisor: ZHANG Qian / CSE Student: PENG Zhuoxuan / COSC Course: UROP1100, Spring Dynamic hand gesture recognition is a classic classification task in computer vision. However, since operations on videos usually occupy a large amount of computational resources, it is difficult to implement a recognition model on Raspberry Pi due to the limited performance. In this paper, I propose a method for monitoring and recognizing dynamic hand gestures on Raspberry Pi using a detector to identify the occurrence of gestures and a classifier to differentiate various gestures. To minimize the computation, the classifier is implemented with an efficient model, 3D ShuffleNet. A large scale dynamic gesture dataset, 20BN-Jester dataset is utilized to pre-train the model on a high performance computer, then the model is transferred to the data recorded with the Raspberry Pi camera module. The recognition accuracy reaches 85.1% on the validation set of 20BN-Jester and 73.3% on data recorded with Pi camera. A single computation time for each sample is 1.39 second. School of Engineering Department of Electronic & Computer Engineering 96 Department of Electronic & Computer Engineering Projects in Audio Signal Processing - Advanced Automobile Crash Detection by Acoustic Methods Supervisor: CHAU Kevin / ECE Student: CHEN Yijia / COSC SIM Yi Hang / EEGBM W AN Yuxuan / CPEG Course: UROP1100, Fall UROP2100, Spring UROP2100, Fall UROP3100, Summer UROP3100, Fall Modern automobiles are equipped with a variety of safety features, but have yet to widely utilize acoustics and sound, although these can provide valuable information and advance warning of a potential car crash. Research was conducted on a method to utilize acoustic waves from a car or surrounding cars, to predict in real time a car crash in advance. The algorithm uses the discrete wavelet transform, which provides multiple time and frequency resolution and a different perspective on sound. By weighting different components of the sound, events such as honking, tire skidding, and collision can be singled out. In particular, tire skidding can be used to predict a car crash and provide information to a car, which can then activate other safety features such as hazard lights or seatbelt tensioners. Overall, the algorithm performs well and has advantages in sudden onset detection, temporal localization accuracy, and computational cost. K. Y. H. Sim, Y. Chen, Y. Wan, and K. Chau, “Advanced Automobile Crash Detection by Acoustic Methods,” presented at the Acoustical Society of America 178th Meeting, San Diego, California, USA, Dec. 2019. Proceedings of Meetings on Acoustics, vol. 39, 055001 (2019). https://asa.scitation.org/doi/abs/10.1121/2.0001150 School of Engineering Department of Electronic & Computer Engineering 97 Projects in Audio Signal Processing - A Real-Time Wavelet-Based Algorithm for Improving Speech Intelligibility Supervisor: CHAU Kevin / ECE Student: CHEN Yijia / COSC SIM Yi Hang / EEGBM WAN Yuxuan / CPEG ZHENG Jiakun / ISDN Course: UROP1100, Fall UROP2100, Spring UROP2100, Fall UROP3100, Summer UROP3100, Fall UROP1000, Summer Presently reported is a wavelet-based algorithm to improve speech intelligibility. The speech signal is split into frequency sub-bands via a multi-level discrete wavelet transform. Various gains (or attenuations) are applied to the sub-band signals before the signals are recombined to form a modified version of the speech signal. Dynamic range compression then follows to control the peak amplitude. The sub-band gains are adjusted while keeping the overall signal energy unchanged, and the speech intelligibility under simulated hearing loss conditions and various background interferences such as babble, music and machine noises is evaluated objectively and quantitatively using Google Speech-to-Text transcription. For both English and Chinese speech, intelligibility is enhanced, and the transcription accuracy can increase by over 60 percentage points by reallocating the audio energy from low to mid and high frequency sub-bands, effectively increasing the consonant-to-vowel intensity ratio. This is reasonable since the consonants are relatively weak and of short duration, and are therefore the most likely to become indistinguishable in the presence of background disturbances or high frequency hearing impairments. The proposed algorithm is implementable in real-time and simpler than others. Potential applications include speech transmission, hearing aids, machine listening, and a better understanding of speech intelligibility. Y. Chen, Y. Wan, K. Y. H. Sim, J. Zheng, T. Kang, E. Chau, and K. Chau, “A Real-Time Wavelet-Based Algorithm for Improving Speech Intelligibility,” to be presented at the Acoustical Society of America 179th Meeting, Dec. 2020. Semiconductor Nanowire-Based Sensors and Electronics Supervisor: FAN Zhiyong / ECE Student: CHEN Runzhou / ELEC Course: UROP1100, Summer In this work, we did research on a high-performance semiconductor nanowire-based gas sensor consisting of a multiplexed tin oxide nanotube sensor array. At the beginning we studied the attribute of target gas and the structure of the sensor and then discussed how the high-sensitivity gas detection and discrimination have been achieved at room temperature. Secondly, we conducted research on the sensor’s response to different baleful gas including H2S, NO2, CO, O2, H2, methanol, ethanol, acetone, etc. Mass flow controller (MFC) ventilation system is used to provide cycles of gas. The ADC voltage is stored. Lastly, the result graph will be plotted to demonstrate the sensor’s response to gases. Accordingly, we calculate related parameters and analyze the result. School of Engineering Department of Electronic & Computer Engineering 98 AI Art Supervisor: FUNG Pascale / ECE Student: CHENG Yih / ELEC ZHANG Yao / COMP Course: UROP1100, Spring UROP1100, Spring This UROP project is to construct a website which implements two different models of style transfer art. According to Pascale FUNG, our supervisor, the purpose of this UROP project is to create art, especially dealing with images. We read papers and searched for references online for background knowledge and will also make use of Github projects to complete our website. We learned about the basic concept about image style transfer and also about Tensorflow machine learning. This was a tough semester due to COVID-19, which forced us to switch to online lectures, however we managed to get to campus and joined the weekly meetings with Professor FUNG and PG students. We learned a lot from the meetings and enjoyed our time finishing this AI Art UROP project. Fully Automatic Movie Generation Using AI Supervisor: FUNG Pascale / ECE Student: CHENG I-tsun / COMP Course: UROP1100, Spring UROP2100, Summer Automatic movie generation is a challenge to AI that has not yet been done before. We introduce a completely novel approach to fully automatic movie generation using a combination of transfer learning techniques, transformer-based models, video-captioning systems, and retrieval methods. We present four primary modules that produce decent results: Plot Generation, Video Captioning, Video Retrieval, and Conversation Generation. Plot Generation utilizes finetuned transformer and Bag-of-Words classifier, Video Captioning adopts information from both visual and audio modalities, Video Retrieval applies human and gender detection, and Conversation Generation leverages pretrained transformer for a sequence-to-sequence task. The end result combining all of the modules together is a short film that is quite relevant to the input genre and also decent to the audience. FPGA-Controlled Silicon Photonic Switches for Datacenters Supervisor: POON Andrew Wing On / ECE Student: TANUWIJAYA Randy Stefan / PHYS Course: UROP1100, Fall The objective of this project is to improve response time and increase the efficiency of the real-time feedback control system for microring resonators in silicon photonics chips functional for transmission channel switching in datacenters regardless of temperature or external disturbances. This report investigates the correlation between resonance wavelength of resonators and the value shift due to applied electrical signals in Thermo-Optical or Electro-Optical tuners by analyzing the photocurrent. Referring to the data, refinement of hardware interface for signals transmittance was performed by revising utilized components for accuracy and precision in addition to the prevention of short-circuit issues. Furthermore, the development of finite-state-machine (FSM) programmed in the Zedboard Field-Programmable-Gate-Array (FPGA) device would enhance the signal processing and hence the photonics switches stabilization. School of Engineering Department of Electronic & Computer Engineering 99 Automated Test Equipment for Magnetic Memory and Sensors Supervisor: SHAO Qiming / ECE Student: CHAN Yin Pui / CPEG Course: UROP1000, Summer There is a lot of instruments and machines involving in a single experiment. Previously, people may control all the hardware by a chunk of lengthy program with low flexibility. It is difficult to manage different servers and do data analysis when it comes to further development and debugging. This project aims to set up a demo of automated test equipment is build using LabRAD. LabRAD allows to break up a complicated system into tiny and simple modules and set ups a network connection among all modules, making the whole system more manageable. The demo is set up to mimic a real-life situation to collect data from servers and analysis the measured data. The demo should be able apply in real-life situation after customizing some servers and client. Equipment Automation for Remote Experimentation Supervisor: SHAO Qiming / ECE Student: LV Zhi Yuan / CPEG Course: UROP1000, Summer In this UROP program, I worked with teammates with the target of building a LabRAD system, which integrates the controlling of all kinds of instruments into one platform. Labrad is not a popular platform, which is unfamiliar to most of us. Firstly, we study the LabRAD wiki to figure out how to use the components of the system, like manager, explorer, and grapher. Then, from the files on GitHub, we learned the grammar of LabRAD servers and clients. To achieve the objectives, we mainly use python to write the code with the help of pylabrad package. Equipment Automation for Remote Experimentation Supervisor: SHAO Qiming / ECE Student: TSANG Yik Wai / SENG Course: UROP1000, Summer The name of this project is ‘equipment automation for remote experimentation’. The purpose of this project is to create a universal software platform to control all the experimental devices remotely. During the project period, different small tasks were assigned to me by the supervisor to help accomplishing the goal. Due to covid-19, those tasks mainly involved python programming. More precisely, the objective of those tasks is to study the performance of a tool called ‘Labrad manager’ which is a system that enable us to control and retrieve data from the experimental apparatuses remotely. In this project, pylabrad which is the python version of labrad is used. Rational Design of Magnetic Devices Using Micromagnetic Simulations Supervisor: SHAO Qiming / ECE Student: CHEN Zehan / PHYS FONG Cheuk Pan / PHYS Course: UROP1100, Fall UROP1100, Fall Skyrmion, a topological object discovered in ferromagnetic systems, has been considered as a potential information carrier for high-density and non-volatile memory. Simulation has been proved an effective approach to study skyrmion motion and deformation in materials with different physical parameters, which also guides experimental and theoretical studies. Utilizing simulation to study skyrmion motion and deformation is crucial for stable and efficient skyrmion transportation. By using other analytical tools such as python, the simulation results are well understood. School of Engineering Department of Electronic & Computer Engineering 100 Rational Design of Magnetic Devices Using Micromagnetic Simulations Supervisor: SHAO Qiming / ECE Student: CHEN Zehan / PHYS Course: UROP1000, Summer UROP2100, Spring Skyrmion is a topological object in ferromagnetic systems, promising for future high-density and non-volatile memory, where the skyrmion dynamics is a major concern. As a spintronic object, skyrmion can be driven by current-induced spin orbit torque (SOT), which will also cause the skyrmion to deform. This report will discuss conditions and patterns of skyrmion dynamics with the help of micromagnetic simulation. Rational Design of Magnetic Devices Using Micromagnetic Simulations Supervisor: SHAO Qiming / ECE Student: FONG Cheuk Pan / PHYS Course: UROP2100, Spring Magnetic skyrmion, a topological, quasi-particle in magnetic materials, is a potential information carrier for memory and computing devices. Its nucleation, stability, dynamic properties were investigated intensively. Both theoretical, experimental studies and magnetic simulation on effect of different parameters on size of skyrmion were taken. Some parameters such as exchange energy, magnetic anisotropy, Dzyaloshinskii–Moriya interaction, and magnetic field can affect the size of skyrmion. Nevertheless, current-induced skyrmion deformation is rarely studied. In this report, magnetic simulation was proceeded for understanding the relationship among DMI, saturation magnetization, Gilbert damping constant, and skyrmion size under different values of current density. The objective is to conclude a general way to determine the size of skyrmion. Rational Design of Magnetic Devices Using Micromagnetic Simulations Supervisor: SHAO Qiming / ECE Student: NG Isaac / PHYS Course: UROP1000, Summer The damping parameter in ferrimagnets is thought to be strongly temperature dependent and strongly increase at the angular momentum compensation temperature. New theory and experimental evidence suggest that the damping parameter is independent of temperature around the angular momentum compensation temperature. To further understand the temperature dependence of the damping parameter in ferrimagnets, we analyze several data sets from literature with this new theory. Using three different methods to estimate the individual sublattice magnetization, we found that the corrected damping parameters do not increase at the angular compensation temperature and support the new theory. School of Engineering Department of Electronic & Computer Engineering 101 Multi-Agent System Control Supervisor: SHI Ling / ECE Student: FU Zhengyu / ISDN XU Jinyun / ELEC Course: UROP1000, Summer UROP1000, Summer A multi-agent system is a robotic system composed of multiple identical robots. It is commonly used to solve problems that are difficult or impossible for an individual agent to solve. In this research, we mainly focus on the formation control of multiple wheeled robots with an overhead camera serving as a global observer. At the current stage, we have finished developing many essential building blocks, including the velocity control, position control, and localization of a single robot, and set up a simulation environment in the gazebo. We also finished evaluating the stability of inter-agent communication through Wi-Fi, and the result met our expectations. Our next step is to investigate the collective behavior of such a system and deploy a distributed control algorithm to each agent to work as a cluster. Attributes Evaluation for ALS Disease Progression Supervisor: WANG Yiwen / ECE Student: LI Xiang / CPEG Course: UROP2100, Fall This report shows a novel method to detect some important features of ALS with clinical data from DREAM ALS Stratification Prize4Life Challenge. The system includes two steps, feature selection and prediction. Feature selection part using a local learning based algorithm to select important features toward ALS progression and prediction part uses a kernel LMSE algorithm to predict ALS progression and evaluate selected features. We evaluate our features in both survival prediction and progression prediction, find out our features perform better than others’ work. Microchip for Automated Manipulation of Biological Cells through Electric Fields Supervisor: YOBAS Levent / ECE Student: TANG Zili / ELEC Course: UROP2100, Fall UROP3100, Spring In this UROP project, I seek to use AC electroosmosis to explain the particle oscillation in microfluidic channel under AC voltage observed in UROP 2100. Simulation showed that AC electroosmosis could generate circulations inside the channel. The simulated circulation frequency shows perfect matching with experimental data at high AC frequencies but discrepancy at lower AC frequencies. The particle in simulation always become trapped at the corner of the channel, which differs from experiment. I also experimented on channels with vertical sidewalls and did not observe the same oscillations from UROP 2100, which proves that the oscillations are related to AC electroosmosis circulations caused by the distinctive geometry of our channel, instead of interaction between particle and the channel wall. VLC-Reckoning: An Indoor Navigation Application that Combines Visible Light Communication and Dead Reckoning Supervisor: YUE Chik Patrick / ECE Student: CHEN Runzhou / ELEC Course: UROP1100, Fall Pedestrian Dead Reckoning (PDR) is used widely for indoor localization where GPS cannot work precisely. The main advantage of PDR is that it does not require any complicated device, just based on the build-in sensors of smart phone. However, there are limitations such as the error accumulation, white noises and complicated body movements. Regarding to these issues, we would combine visible light communication (VLC) with PDR to increase the accuracy. School of Engineering Department of Industrial Engineering & Decision Analytics 102 Department of Industrial Engineering & Decision Analytics Big Data Analytics Techniques for Quality and Process Improvement Supervisor: TSUNG Fugee / IEDA Student: CHOU Chih-an / COMP Course: UROP1100, Fall Customer Satisfaction Index (CSI) is an indicator which measures how satisfied the customers are. This study was conducted to determine the customer satisfaction and the factors which influence it of the public transportation system in Hong Kong only based on the online reviews. This paper focuses more on Mass Transit Railway (MTR) than all the other public transportations in Hong Kong, and was performed on 17,201 online comments collected from TripAdvisor.com and written by MTR users from 2015 to 2019. The reviews were all analyzed in an unsupervised way since there was no label given to the reviews. More specifically, the online reviews were examined through Latent Dirichlet Allocation (LDA) model, Valence Aware Dictionary and sEntiment Reasoner (VADER), and K-means Clustering. The result of the analysis presents that the customer satisfaction of MTR in Hong Kong is overall positive and the factors extracted from the analysis which will critically affect customer satisfaction toward public transportation are travel time, accessibility, cleanliness, ticket fare, safety, customer service and crowdedness. It appears that the better the factors listed above perform, the higher the customer satisfaction of the public transportation will be. Big Data Analytics Techniques for Quality and Process Improvement Supervisor: TSUNG Fugee / IEDA Student: WANG Xiaojie / IEEM Course: UROP1100, Fall An app named Octoparse was used for its convenience and low requirement on programming language. The target website is set to be Hotels.com for two reasons: It provides sufficient amount of reviews with a wide time range and the websites structure is suitable for Octoparse. Octoparse only provides very basic logic operations for data extraction, so it is impossible to extract reviews from most of the websites. So the process needs to be carefully studied and modified to fit Hotels.com. (e.g. the loop with click item and turn next page is not provided by Octoparse) Big Data Analytics Techniques for Quality and Process Improvement Supervisor: TSUNG Fugee / IEDA Student: ZHANG Yunfei / SENG Course: UROP1100, Fall Service is one of the most significant dimensions that is focused when people are choosing hotels. Feedbacks from the customers are often collected by hotels after they offer the service. Most of the time, these feedbacks can reflect the true fresh service level of the related hotel. Then the hotel can make decisions to offer a better service next time. Nowadays, with the development of internet, people can write the reviews through many different stages on the internet. Every day passing, such kind of reviews accumulate a lot as a kind of text data. For hotels, they want to further develop their service level by analyzing these review data, and this is exactly what this project focus on – online text data analysis by models. School of Engineering Division of Integrative Systems & Design 103 Division of Integrative Systems & Design Gait Analysis for Control and Optimization of Exoskeletons Supervisor: LEUNG Suk Wai Winnie / ISD Student: SYED ALI Mohamed Fouzil Ali / CPEG Course: UROP1100, Fall This UROP 1100 report demonstrates the applicability of a deep reinforcement learning based approach for the gait analysis human locomotion and proposes it as an alternative option to be used in the HKSUT-Cybathlon Exoskeleton project. The conclusion was arrived after simulating existing control theory based modelling and then applying a deep reinforcement learning approach to the same simulation to produce similar results but with the added benefits of having the potential to take advantage of the non-linear function approximation offered by neural networks. The report establishes a comprehensive metric to evaluate any approach used to evaluate human locomotion gait and brings out the deficiency in the current implementation after a brief explanation as well as a proof-of -concept implementation of the alternative deep reinforcement learning approach. Terrain Sensing and Mapping for Exoskeleton Design Supervisor: LEUNG Suk Wai Winnie / ISD Student: RASTOGI Tanya / COMP Course: UROP1100, Fall State of the art technical systems are used by physically disabled people to aid them in executing day to day tasks. One competition that focuses on this technical creativity is the Cybathalon. For a technical system to aid people with lumbar spinal cord injury or complete thoracic injury, it needs to accurately map the surroundings for the pilot to walk and clear obstacles without harm. The initial step of this mapping includes depth recognition while walking. The distance of the foot from the step or ground and how and where it should be placed is based on the depth of the objects around the foot. This paper analyses the three most optimal methods for depth sensing: Structural light, Time of Flight and Camera Array. It compares the three methods based on cost and overall performance to sense depth and map the surroundings accurately. Development of a User-Friendly, Efficient Management System for Stair Climbing Service Supervisor: LING Carrie / ISD Student: CHENG Log G / COGBM WANG Yufei / COGBM Course: UROP1100, Fall UROP1100, Fall This project aims to develop a user-friendly, efficient booking and management system for the stair-climbing service offered by St. James’ Settlement Continuing Care by integrating Google services on top of St. James’ existing website as well as other booking management infrastructure. This method has several advantages over their existing system: Firstly, it is more maintainable and scalable, since designed documents are readily available and constantly maintained by Google. Secondly, it is more economical, as the services provided by Google are mostly free. Within this semester, we have migrated their original booking system on their website to Google’s services, including the booking form, as well as the back-end booking management system. School of Engineering Division of Integrative Systems & Design 104 3D Scene Generation with Scanned RGB-D Data Supervisor: YEUNG Sai Kit / ISD Student: IEONG Zi Liang Jason / COMP Course: UROP1100, Spring Our target is to submit a conference paper for CVPR 2021 with a title of ”Realtime Scan2CAD” as well as building a robotics framework that allows the robot to conduct visual Simultaneous localization and mapping (vSLAM), and identify CAD object within the database and retrieve it in the physical world at the same time. The goal within the period of this UROP project is to learn the required basis and built the required environment for the final framework. This UROP progress report will show my self-learning progress on machine learning, vSLAM and the Robot Operating System (ROS2). 3D Scene Generation with Scanned RGB-D Data Supervisor: YEUNG Sai Kit / ISD Student: PANG Hong Wing / COMP Course: UROP1100, Fall UROP2100, Spring Image generation via artificial intelligence is a widely explored topic over the past few years. The idea of a machine able to generate realistic unseen images is not only intriguing, but also opens up many possibilities including computer vision and graphical applications. In particular, we are interested in the generation of indoor scene images, such as photos of interior designs of buildings and apartments; the ability of automating the generation of such images would simplify the design process of apartment flats and provide a simpler way of visualizing the arrangements of furnitures and lighting, instead of relying on procedural rendering methods. The generation of indoor scene objects has been attempted before, but the performance of existing works remain far from realistic compared with the generation of other types of images. Hence, our goal is to improve the generation quality of existing methods on indoor scene images. 3D Scene Generation with Scanned RGB-D Data Supervisor: YEUNG Sai Kit / ISD Student: ZHANG Yukai / ISDN Course: UROP2100, Fall UROP3100, Spring UROP4100, Summer Point cloud—a set of unordered points in 3D space, is a common geometric data structure typically generated with the use of 3D laser scanners. As one of the simplest ways to represent 3D structures, many 3D deep learning methods focus on point clouds. However, most of today’s point cloud datasets for machine learning, such as ModelNet40 and ShapeNet, are derived from CAD models. Although very high performances have been achieved on these datasets with many classification and segmentation methods, these tasks can still be challenging when the point clouds are realworld scans. Previously, we have experimented the classification performance of various neural networks on the realworld scan dataset ScanObjectNN. In this project, we cleaned and processed the part segmentation data in ScanObjectNN and tested the performance of PointNet and PointNet++ on part segmentation. School of Engineering Department of Mechanical & Aerospace Engineering 105 Department of Mechanical & Aerospace Engineering Investigating the Electrodeposition of Zinc from Alkaline Solutions Supervisor: CHEN Qing / MAE Student: CHAN Cheuk Kai Kevin / SSCI Course: UROP1100, Summer The development of zinc anode rechargeable cells is still facing many problems. One main problem is the morphological change of Zn during electrodeposition. The formation of microstructures of Zn, such as mossy structure or dendritic structure, may cause a fatal operation of Zn batteries. Therefore, to find an optimal condition for the electrodeposition of Zn, it is worthwhile to investigate the morphological change of Zn under different conditions. In this report, the morphological and microstructural change of Zn under different current densities, overpotentials, electrolytes and additives will be discussed. Design of Energy Conversion Device from Heat to Electricity Supervisor: CHEN Sherry / MAE Student: SUEN Ho Yuen Henry / MAE Z HANG Yuchen / AE Course: UROP4100, Fall UROP1100, Fall The following report presents the design of a solar heating prototype which reflects solar irradiation and focus it to rapidly heats up a small slice ferroelectric material. This report is both a group report and a progress report. SUEN, Ho Yuen Henry focused on CAD drawing and online orders. Zhang Yuchen conducted most of calculation for design parameters. Through this UROP project period we have mainly finished the design parts and made final purchasing. However, future test on the design is still needed and may include change on design. Power-Source-Free Energy Generator Supervisor: CHEN Sherry / MAE Student: ZHANG Yuchen / AE Course: UROP1100, Spring UROP2100, Summer This summer we mainly focused on parameters measurement and raw material selection for our project. Rather than theoretical prediction, we conducted an estimative measurement on CPU temperature variation and obtained abundant data to facilitate the design of energy conversion module. Simultaneously, literature research has been done and we found a kind of special polyethylene, which has a property of unidirectional thermal conduction, may be an ideal material to manufacture thermal conductor for our module. Yet there are still problems need to be solved. Since the computer we used has temperature protection mechanism, integrated cooling devices cannot be removed so the experiment could not accurately reflect CPU’s thermal properties. To rectify the current error, a simulative approach will be adopted later on. School of Engineering Department of Mechanical & Aerospace Engineering 106 Multidimensional Modeling of Nanoionics Supervisor: CIUCCI Francesco / MAE Student: DELA CRUZ Aragorn Joaquin Pineda / SENG Course: UROP1000, Summer The field of nanoionics holds considerable potential and widespread applications in energy storage and conversion devices including fuel cells and batteries, yet still lacks conducive study on impedance responses of multidimensional, nanosized systems. This article pertains to the utilisation of computational models and algorithms to investigate the alternate current response of nanoionic systems. Specifically, an analysis of the accuracy of an algorithm designed to calculate the Hilbert transform of a function based on the rational eigenfunction expansion method was carried out. It was found that the method produces especially comparable results with residuals in the range of 10−2 to 10−17 , and various parameters can have a moderate effect on the accuracy of the algorithm. Polymer Composite Batteries Supervisor: CIUCCI Francesco / MAE Student: YU Tsz Tat / MAE Course: UROP1100, Fall UROP2100, Spring UROP3100, Summer Secondary lithium-ion batteries with high energy density and long cycle life are indispensable to the advent of sustainable energy and transport. The development of solid-state electrolyte alongside with high capacity electrodes are critical to the performance gains. Based on the dual-salt polymer electrolyte developed in the previous project, three new approaches have been proposed to improve the battery performances, including adding perovskite solid electrolyte, replacing electrolyte solvent with fluorinated 1,4-dimethoxylbutane (FDMB), and replacing metallic lithium anodes with silver-carbon nanocomposite anodes. These approaches will be experimented in the future. The group’s work provides a clear path towards energy-dense, long-living, and safe solid-state Li-ion batteries. Application of AI-based Technique to Enhance Thermal Comfort Sensing for Smart Air Conditioner Supervisor: LEE Yi-Kuen / MAE Student: SHE Fong Wing / RMBI Course: UROP1000, Summer In this report, it includes the summary of learning and the details about the machine learning model for human activity classification in process during the summer semester of UROP 1000 course. The summary of learning includes the detail explanation of different concepts learnt, particular Artificial Intelligence, Air Conditioner, Predicted Mean Vote, MIT App Inventor, Machine Learning Algorithms, Neural Network, Computer Vision, and Recurrent Neural Network. For the machine learning model for human activity classification in process, data collection using the accelerometer and gyroscope in Android phone and data analysis using Google Colaborotary and Matplotlib Library are done with the model building remaining in process. School of Engineering Department of Mechanical & Aerospace Engineering 107 Application of Artificial Intelligence to Enhance the Fluorescence Microscopy of Circulation Tumor Cells Captured by MEF Chips Supervisor: LEE Yi-Kuen / MAE Co-supervisor: ZHAO Cong / BME Student: LEE Khang Sek / MAE Course: UROP1000, Summer The objective of this research is to explore the application of Artificial Intelligence in detecting circulating tumor cells (CTCs), which serves as an alternative for conventional cancer diagnostics that is often expensive and time-consuming. It is understood that within the scope of Artificial Intelligence, Deep Convolutional Neural Network (DCNN) stands out to be the best solution to perform image classification task as required in our research to identify CTCs. To apply DCNN algorithm, I started from learning the programming language Python, then followed by learning Google TensorFlow, a core open-source library to build and train machine learning model before picking up Keras, which is a high-level API that uses TensorFlow as backend. Combining all I have learned, I managed to reproduce the results of skin cancer cell detection using DCNN and achieved 80% categorical accuracy using 10,500 open-source online skin lesion images. Development of Low-cost Smartphone-based Microscope for Global Health Supervisor: LEE Yi-Kuen / MAE Student: JUNG Minyoung / BIOT Course: UROP1000, Summer Smartphone industry has been making good progress in developing new smartphones with advanced CMOS image sensors and the other MEMS sensors in recent years. More people around the world can access smartphones and take high-quality pictures with the integrated camera. This indicates the potential of developing low-cost, portable microscope that can be clipped on to the smartphones for infectious disease diagnostic and monitoring purposes, especially in the developing regions that lack infrastructure and professionals who use the laboratory microscopes. This report studies the mechanism of a microscope, application of microscopy in the global health setting, and reviews currently existing and proposed models for smartphone integrated microscopes. Furthermore, the report proposes a student-designed microscope prototype using SolidWorks based upon the existing models. Low-cost Wireless Arduino-based Micro Energy Sensor for Smart Energy-Efficient Buildings in the era of Internet of Things Supervisor: LEE Yi-Kuen / MAE Student: YUEN Zhikun / COMP Course: UROP1000, Summer In this project, we need to design a way to build a system for saving the energy by controlling the devices of a building. The Arduino board is a controller which control all the IOT devices inside a building like air conditioner or light, etc. Normally, we may use machine learning to train a model and use it to help us control the usage and save energy. However, we found that simply algorithms and formulas can replace the machine learning method and make it simpler. As a CSE student, I helped the group to build a mobile app to communicate with an Arduino board through Bluetooth and implement the formulas in the app. School of Engineering Department of Mechanical & Aerospace Engineering 108 Development of Extended-Range Unmanned Surface Vehicles (USV) Supervisor: LI Larry / MAE Student: JACOB Jubin / AE Course: UROP1100, Fall Owing to its diverse applications and advantages over its manned counterparts, the USV market size is projected to cross a billion US dollars by 2023. Unmanned /Autonomous Surface Vehicles are vessels designed for operation on water surfaces without the aid of a crew. The redundancy of onboard crew enables USV layouts to be streamlined with respect to their objectives in a cost-effective manner. Typical USV applications include ISR (Intelligence, Surveillance, Reconnaissance) missions, oceanology investigation, water quality monitoring, bathymetry etc. Although the lack of crew onboard permits USVs to remain offshore for longer periods, the duration of operations is constrained by the limits of conventional energy storage systems. The study explores the feasibility of renewable energy with special emphasis on wave energy as a means to improve operation times through literature review. Development of Extended-Range Unmanned Surface Vehicles (USV) Supervisor: LI Larry / MAE Student: KIMISHIMA Hiromi / AE Course: UROP1100, Fall Energy harvesting for Unmanned Surface Vehicles (USVs) is a key issue that has to be solved for a widespread usage of USVs. In addition to common renewable energy sources like wind and solar energy, USVs have access to oceanic waves, which has a higher energy density than wind or solar energy. However, wave energy is not as prevalent as wind or solar energy, and harvesting is still limited. This report summarizes the past literature regarding wave energy harvesting methods and also considers the potential for an energy harvesting method for USVs. Nonlinear Dynamics and Closed-Loop Control of Droplet Formation in a Capillary Jet Supervisor: LI Larry / MAE Student: OLSSON Christopher Sven / AE Course: UROP1100, Spring UROP2100, Summer A brief review of the currently available literature on propeller tip design and commercial UAV propeller design. There is currently no stand-alone metric in quantifying the efficiency of a propeller tip (energy and thrust lost). Low Reynold (Less than 50,000) propeller tip effect is increasingly important due to expanding usage of micro-drone. Therefore propeller tip investigations could yield interesting insight. School of Engineering Department of Mechanical & Aerospace Engineering 109 Aircraft Design and Air Transportation Operations: How Do They Affect Each Other? Supervisor: LIEM Rhea Patricia / MAE Student: CHAVANACHINDA Phisitchai / SENG SWAMINATHAN Srivatsav / SENG YIU Kwan Ho / SENG Course: UROP1000, Summer UROP1000, Summer UROP1000, Summer The COVID-19 pandemic has posed a significant impact on air travel. The sharp decline in demand has driven airlines to reconsider their operational strategies and has compelled both airlines and aircraft manufacturers to find ways to optimize costs. This paper aims to analyze the trend in air transportation models, to examine aircraft design that could potentially optimize costs, and to identify the most efficient aircraft types for specific service ranges. This serves to help establish a link between aircraft design and air transportation operations, which are often disconnected from each other. The results show that larger aircraft are more cost-efficient when fully occupied, among other factors, supporting the analysis of a shift in the model of air transportation operations due to the pandemic. School of Engineering Department of Mechanical & Aerospace Engineering 110 Water Tank Experiment Setup and Integration for Amphibious Aircraft Design Research (HKUST and SJTU) Supervisor: LIEM Rhea Patricia / MAE Student: FONG Wai Ting / AE LIU Kai Kong / MAE SHIHUA Mingzhi / AE Course: UROP1100, Spring UROP1100, Spring UROP1100, Spring This project serves as a continuation of the previous Amphibious Aircraft UROP by researching effects of ground effects on amphibious aircrafts and by building a database which relates take-off performance and its relevant parameters.2D Computational fluid dynamics (CFD) involving different airfoil shapes and ground clearance were performed. The results suggested that ground effect would not produce any noticeable changes on aerodynamic performance. Regression analyses were performed to create low-fidelity models of resistance with respect to speed and load. Two surrogate modelling methods were compared, namely RMTB and Kriging, and the model parameters were tuned. We conclude that the RMTB method is better than Kriging method in this case. School of Engineering Department of Mechanical & Aerospace Engineering 111 Biomimetic Design of Flexible Flapping Wings for Micro Air Vehicles Supervisor: QIU Huihe / MAE Student: SUN Ka Ip Justin Edmund / MEGBM Course: UROP2100, Fall UROP3100, Spring The flapping flight of insects and birds is more complex involving unsteady aerodynamics, separated flow and transient effects, unlike traditional fixed-wing aerodynamics. In this project, the focus has been on wing kinematic effects of Pantala Flavescens, a dragonfly species for the bioinspired design of Micro Air Vehicles (MAVs). Specifically, the contribution of active, muscle produced pitching and the passive pitching resulting from the flexible membrane deformation of the wing in response to aerodynamic has not been well understood. This project is to collect data and analyse the flight kinematics of two different flight modes observed: “Normal mode” and “Escape mode”. To achieve this, high-speed cameras were used to track certain points of the dragonfly’s wings, to calculate the pitching angles, positional angles, phase differences and flapping frequency while simultaneously tracking the force it exerts to generate lift and thrust to characterise the flight modes, thus helping to build a model for flapping flight for MAVs. Biomimetic Design of Flexible Flapping Wings for Micro Air Vehicles Supervisor: QIU Huihe / MAE Student: YEUNG Hon Kit / MAE Course: UROP1100, Spring Micro air vehicle (MAV) is useful in the military, for example, monitoring target. To complete it, it is necessary to study the flying mechanisms of birds and insects, and in this project, it is mainly focusing on the flying mechanisms of insect. And the flying mechanisms can be affected by its size, and the structure of its wings, for example, the number of its wings and is it allowing the air to pass through it. And the wings movement during hovering, lifting, landing, accelerating, decelerating, and turning will also be studied. It is also required to know how humidity, temperature, and airflow affect the flying mechanism since MAV is expected to use in different environments. [Control Engineering] Underwater Robot for Maritime Exploration Supervisor: SEO Jungwon / MAE Student: MAK Ka Hei / MAE Course: UROP3100, Fall This report summarizes the progress of designing an omni-directional robot for rock-walk manipulation. An omnidirectional robot was built with the caging end-effector to transport and steer a circular cone object. The design goal of the omni-directional robot is to perform an accurate moving speed in X, Y and rotation axis with the desire target velocity. With an accurate response of moving speed in 3 axes, the robot with a caging end-effector can manipulate the cone object by moving in a zigzag path. The prototype of the omni-directional robot shows its capability of making a successful manipulation with the control of a gaming controller. However, the property of low frictional force from the omni wheels causes the robot to slip easily in high acceleration, in which leads to an inaccurate moving path. Several improvements for optimizing the robot performance will be made in the future. School of Engineering Department of Mechanical & Aerospace Engineering 112 Dexterous Manipulation with Soft Robotic Gripper Supervisor: SEO Jungwon / MAE Student: REN Jiming / MAE Course: UROP3100, Fall The energy approach we have developed in our flex-and-flip maneuver is general and well applicable on various grippers. In flexflip theory, we generate bending energy of a unit length of paper and the minimum coefficient of friction required at the contact point to keep the curve of deformed paper steady in various configurations. This model perfectly explains both the desired and undesired scenarios presented in experiments through examining a critical instant when the flexural energy of paper is large enough in a direction to overcome the constraint of the friction, lose contact with the fingers and get enough space to be fully flipped over. In our case, we apply Yale T42 to implement the same maneuver to prove the universality of the theory. Robotics Gripper Development for Warehouse Automation Supervisor: WANG Michael Yu / MAE Student: LEE Man Yeung / MAE Course: UROP1100, Spring Gecko is known for its agile and reliable climbing ability. Scientists and engineers have studied the underlying mechanism for many years, in order to make use of this special feature to realize controllable gecko-inspired adhesive applications in the manufacturing industry, aerospace, or our daily life. This paper reviews the mechanism and development of common gecko-inspired dry adhesive surfaces. Then, this paper summarizes the pros and cons, and potential solutions regarding the characteristics of the existing materials. Further, the recent applications of the geckoinspired dry adhesive materials are reviewed. Finally, the prospect and focus of future development are discussed, and other potential applications will be introduced. Robotics Gripper Development for Warehouse Automation Supervisor: WANG Michael Yu / MAE Student: LIANG Yuchen / ISDN Course: UROP1100, Summer This UROP project is named as Robotics Gripper Development for Warehouse Automation. The naming involves Gripper and Automation which means the research focus on developing a gripper mechanism which can be applied to warehouse environment mainly for automation purpose. The gripper mentioned above is not the normal mechanical gripper we see in our daily life but another kind of special gripper that require special material design for the touching surface of the gripper and the object to be gripped. The material is a kind of gecko inspired dry adhesive which was inspired by the gecko’s feet. For 2020 summer, the UROP research focus on mostly material design and fabrication and the mechanism mechanical design of the gripper. Abstracts of UROP Projects 2019-20 SCHOOL OF BUSINESS & MANAGEMENT School of Business & Management Department of Accounting 114 Department of Accounting Angry Shareholders and Failed Board of Directors Supervisor: CHEN Szu Fan / ACCT Student: BIEK Thomas Michael / BIBU Course: UROP1000, Summer Determining what makes a failed board of directors is key to providing meaningful advice for numerous professional services companies. The shareholder derivative lawsuit is a versatile and easily comparable method of identifying when these boards of directors fail. One key piece of knowledge is the industry distribution of these suits, which could be incredibly beneficial to business services firms. By comparing these suit distributions, it was found that the manufacturing sector was most at risk, particularly in the pharmaceutical preparations industry. Another interesting discovery was the increase in business services and retail trade suits, at the expense of other industries. For consulting and law firms, this information can help them price their services correctly given a certain industry’s litigation risk. Angry Shareholders and Failed Board of Directors Supervisor: CHEN Szu Fan / ACCT Student: LI Xiangyu / ECOF Course: UROP2100, Fall UROP3100, Summer This paper investigates the impact of accounting-related derivative lawsuits on financial reporting quality. We measure this quality by restatements, auditor change, and yearly profitability change both before and after the lawsuit filing year. Comparing defendant firms sued for accounting reasons and those not for accounting reasons, we use a difference-in-differences approach to identify the effect of accounting-related lawsuits on financial reporting quality. We find that firms sued in accounting-related lawsuits become less likely to restate and more likely to change auditors following the filing of lawsuits, and these changes are robust to the inclusion of the control sample. Our study suggests that accounting-related derivative lawsuits could improve financial reporting quality. Angry Shareholders and Failed Board of Directors Supervisor: CHEN Szu Fan / ACCT Student: LUI Hiu Lum / ACCT Course: UROP1000, Summer This report aims to investigate the change in reasons of derivative lawsuits from 1998 to 2017. Filing reasons of these lawsuits were extracted from financial statements of the companies. Then the data was categorized and analyzed for trends and changes throughout the timeframe. Dominant areas for derivative lawsuits have been identified. And it has been found that the trend within recent years is relatively mild, with the only drastic change happened during the 2003 to 2007 period. These findings may help stakeholders to formulate better policies to protect one’s interests by signalling areas with higher risks and reflecting effectiveness of policies. School of Business & Management Department of Accounting 115 Angry Shareholders and Failed Board of Directors Supervisor: CHEN Szu Fan / ACCT Student: YUAN Xuzhi / FINA Course: UROP1100, Summer As one of the major forms of shareholder litigation, derivative lawsuits often involve high costs but result in low rewards for shareholders. Most suits were dismissed by the courts, and monetary recoveries are often insignificant even among the few cases that reached settlement. The demands of plaintiffs in most suits are only partially satisfied or not satisfied (Griffith & Sean J. 2018). Lawsuits that are costly with little recoveries should not be considered by rational plaintiffs. This is especially important in derivative litigation as plaintiffs in these suits do not obtain direct personal benefits. Even with these disadvantages, a considerable amount of derivative suits is still filed every year. Therefore, we would like to investigate and learn more about this special type of shareholder litigation with this research. Angry Shareholders and Failed Board of Directors Supervisor: CHEN Szu Fan / ACCT Student: ZHONG Minhua / ACCT Course: UROP1000, Summer This research is about the impact of derivative litigation on a company’s corporate governance. We use the derivative lawsuit filed against Xerox in 2002 as a case study. We find that there is a relationship between the filing of derivative lawsuit and board independence, the gender diversity of the board, and firm performance. This study suggests that firms may avoid derivative litigation by increasing the independence and the gender diversity of their boards of directors. Cryptocurrency Return Analysis using Machine Learning Algorithms Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YOU Haifeng / ACCT Student: CHENG Sum Yuet / FINA XU Ziqin / RMBI ZHENG Chak Lam / FINA Course: UROP1100, Spring UROP2100, Spring UROP1100, Spring In this article, we aimed at constructing a cryptocurrency portfolio based on high-dimensional daily exchange and sentiment data, and low-dimensional high-frequency data that have been extracted to fit machine learning algorithms. Five mainstream coins including Bitcoin, Ethereum, Ripple, Litecoin, and Bitcoin Cash were tested due to their market dominance and popularity. According to testing results of above-mentioned different data types and their corresponding models, we concluded that high-dimensional daily exchange data with logistic regression generated higher accuracy in future price prediction. The investment strategy based on this model outperformed the market represented by CCI 30 and an AI-driven cryptocurrency portfolio called CIX 100 by gaining a 20% in the second half of 2019 when the market was down. Besides profit, our cryptocurrency portfolio can also provide investors with market insights by explaining influential factors. School of Business & Management Department of Accounting 116 Cryptocurrency Return Analysis using Machine Learning Algorithms Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YOU Haifeng / ACCT Student: LLAMAS Facundo Martin / WBB TSANG Lok Kan Ethan / QFIN ZHENG Chak Lam / FINA Course: UROP1000, Summer UROP1000, Summer UROP2100, Summer This paper ranks and quantifies the effectiveness of different neural networks and machine learning techniques to predict the future price of Bitcoin (BTC). We built different machine learning and deep learning algorithms to predict the future price of Bitcoin. The selected features include: Bitcoin and 13 other major cryptocurrencies trading data, dominant financial instruments' close price (Stock Indices), and Sentiment Data gathered from Social Media. These models can be further separated into two categories: regression, and classification. Regression models construct a formula to predict the next price of Bitcoin, while classification predicts the direction of the future price change. We tested three traditional machine learning methods, OLS Regression, Logistic Regression, LASSO regression and two deep learning methods named Long Short-Term Memory and Gated Recurrent Unit. Our research concluded that among different models, simple machine learning technics outperform the complicated deep learning models. Cryptocurrency Return Analysis using Machine Learning Algorithms Supervisor: HUANG Allen Hao / ACCT Co-supervisor: YOU Haifeng / ACCT Student: TAN Yin Zhen / FINA XU Ziqin / RMBI Course: UROP1100, Fall UROP1100, Fall With the advancement of blockchain technology and digital assets, cryptocurrencies have become new investment options. For example, multiple cryptocurrency index funds like Crypto20 and Bitwise provide investors with marketoriented investments at a low-cost. Among those indexes and index funds, some construction methodologies are proved to enhance performance and reduce price volatility. Their favorable results have attested that it is possible to further improve the performance by varying the index construction process. In this report, we will introduce new methodologies different in selecting, ranking and weighting to construct a cryptocurrency index. Based on the historical statistical performance and market sentiment affected by social media like Twitter, our index aims to choose qualified cryptocurrencies that may be promisingly profitable in the foreseeable future, in order to offer investors an opportunity that beats the bubble market and achieves a higher return with less volatility. Bank Lending Practices and Financial Covenant Renegotiation Supervisor: LI Xinlei / ACCT Student: CHAN Pui Shing / ACCT CHAN Yin Hei / ACCT Course: UROP1100, Fall UROP1100, Fall This research is conducted to explore the relationship between contract loan amount and the frequency of renegotiations of financial contracts. Contract renegotiation is an important mechanism that exists in the corporate world because it allows the borrowing and lending party to use new information to create an improved agreement that benefits both parties. Renegotiations often result in changes to the amount, maturity, and pricing of the contract, meaning it can have significant impact on a company’s financial health. The results generated from this research is important to both businesses looking to borrow and financial institutions looking to lend as it highlights one potential factor that influence such an important mechanism in the world of money borrowing. By studying more than 400 instances of contract agreements, the results will shed light on whether such a relationship exists and the significance of any possible relationship. School of Business & Management Department of Accounting 117 Bank Lending Practices and Financial Covenant Renegotiation Supervisor: LI Xinlei / ACCT Student: CHOI Nim Tung / ACCT ZHOU Zhuorui / COGBM Course: UROP1100, Fall UROP1100, Fall This research studies whether the duration between renegotiations is affected the financial status of a company. Since the financial status can be indicated by the following variables, leverage ratio and cash rate, we investigate the correlations between the variables with the likelihood of renegotiation. We hypothesize that a high leverage ratio will lead to a higher sensitivity of borrowers to mutual benefit, hence borrowers renegotiate more often. Besides, the lowered cash rate will trigger renegotiation more easily since borrows have higher demand in cash, hence shorter duration. The former is supported by the analysis results, but the data does not show a significant relationship on the latter. Bank Lending Practices and Financial Covenant Renegotiation Supervisor: LI Xinlei / ACCT Student: HUI Yuk Ming / ECOF KU Jia Sien / ECON Course: UROP1100, Fall UROP1100, Fall We study whether the performance pricing feature explains the reduction in renegotiation in credit agreements. Recent empirical evidence shows that pricing grids are used to increase future renegotiations, particularly in borrowerunfavourable conditions. Alternatively, previous literature provided a different rationale--performance based pricing, in other words, pricing grids are used to reduce future renegotiations to stave off necessary re-contracting costs which are usually borne by the initiating party. Subsequently, this benefits both parties. Notably, research findings have also documented that firms start from the high-rate end of the pricing grids with most grids allowing for credit improvements. Consistent with this rationale, we aim to show that pricing grids serve the purpose of reducing future necessary renegotiations, specifically decreasing-interest pricing grids that are effective in borrower-favourable conditions. Likewise, increasing-interest pricing grids serve a similar purpose in borrower-unfavourable conditions. Credit agreements with embedded performance pricing feature are less likely to be renegotiated before the maturity date. Bank Lending Practices and Financial Covenant Renegotiation Supervisor: LI Xinlei / ACCT Student: PAN Ai / SBM ZHANG Han / ECOF Course: UROP1100, Fall UROP1100, Fall Building on incomplete contract theory, this research aims to study whether renegotiation frequency can be affected by firm’s characteristics. The incomplete contract theory suggests that renegotiation is a trade-off between efficiency gains and costs, so we hypothesize that firms with better ability to extract efficiency gains will demand for renegotiation more frequently, and firms that run out of the initial contract term more quickly can also expect more frequent renegotiation. Furthermore, we show that the effects of firm size and profitability on the frequency of renegotiation have a strong interaction effect. School of Business & Management Department of Accounting 118 Investment Analysis with Machine Learning Supervisor: YOU Haifeng / ACCT Student: FRANZINELLI Patrick / WBB Course: UROP1000, Summer During the course of the Undergraduate Research Opportunity Project, a paper authored by professor Haifeng You and Kai Cao was replicated. Machine learning models were used to forecast earnings using financial statement data. The forecasts are benchmarked to a common Random Walk model and several extant models from prior literature. The examined machine learning models are able to deliver more accurate predictions than the naïve Random Walk model and the other benchmark models. Non-linear models show the best results; thus, they are able to uncover high level, non-linear relationships between statement lines and actual earnings. We further compared forecasts generated using balance sheet data and income statement data only. Most of the information comes from the income statement and the balance sheet only adds marginal benefit to forecast accuracy. The forecasts generated with the income statement data only, slightly underperformed the trial with all the features, while the forecasts using balance sheet only performed worse than all of the benchmarks. Investment Analysis with Machine Learning Supervisor: YOU Haifeng / ACCT Student: HUANG Zhao-xin / GBUS Course: UROP1000, Summer Given the complex mechanism in commodity future price determination, this project aims at predicting commodity future price movements using machine learning algorithms. Specifically, we investigate the role of historical price/volume, stock market, and fundamental information in driving the future commodities prices. Chinese commodity market is our initial focus, and in our first attempt in predicting Dalian commodity exchange corn return, time series model is employed. Though not yielding positive result, it sheds light on how to improve the feature selection and algorithm in improving the prediction results. Literature review, data collection and testing, algorithm optimizing were conducted in the project. Investment Analysis with Machine Learning Supervisor: YOU Haifeng / ACCT Student: LU Xiaoyi / ECON Course: UROP1000, Summer In this article, we focus on investment analysis of the commodity futures market and investigate the usefulness of machine learning in returns forecasting. In particular, we will try to capture features that have predictive power for the sign or return itself. We use XGBoost to generate models and test predictability on the main contracts of Shanghai copper futures. At this stage, we also research whether the pattern or curve matches time-series models like ARIMA. In follow-up research, the prediction performance of other machine learning algorithms will be tested and compared. The prediction results will be analyzed from the perspective of macroeconomics to conclude what factors drive returns in the commodity futures market and form quantitative strategies based on the factors. School of Business & Management Office of the Dean of Business & Management 119 Office of the Dean of Business & Management Adoption of Digital Banking and Insurance Services in Hong Kong Supervisor: TAM Kar Yan / DBM Student: CHIU Long Kit / GBUS Course: UROP1100, Summer As the first batch of virtual banks and insurers were granted their license to operate in Hong Kong, the study of adoption of digital banking and insurance services in Hong Kong aims to explore the opinions of the public regarding the trust and confidence on these emerging industries as a part of the revolution of Financial Technology. A survey is drafted and issued to the public to assess the public’s understanding and opinions towards the virtual banking and insurance industry. This progress report includes the drafting of the survey questions and data analysis carried out on a part of the questions based on feedbacks provided by respondents of the survey. Comparing Hong Kong with Other Leading Fintech Centers in the World - A Competitive Analysis Supervisor: TAM Kar Yan / DBM Co-supervisor: LAFON-VINAIS Veronique J A / FINA Student: DONG Yu / QFIN GU Zhaoyi / QFIN XUE Haoze / QFIN Course: UROP1000, Summer UROP1000, Summer UROP1100, Summer We try to examine the potential of Hong Kong as one of the leading fintech centers around the globe and to understand the opportunities and challenges it faces in its positioning. A comprehensive analysis is conducted among Hong Kong and other world-leading fintech hubs (Singapore, London, New York, etc.) on various dimensions such as regulation, society support, and market development. Shenzhen is a special focus because of the Greater Bay Area development. Our main conclusion is that the Hong Kong government gives a lot of support to boost the fintech sector, yet it needs to further cultivate the market and expand its international influence. School of Business & Management Department of Economics 120 Department of Economics Across-Country Wage Compression in Multinationals – Evidence from Brazil Supervisor: LI Xuan / ECON Student: LAM Tsz Kwan / ECOF Course: UROP1100, Spring Many multinationals employers may react to the shocks to uncertain risk from policies in their home countries, such as minimum wages and exchange rates fluctuations by adjusting wages or firing employees in branches aboard. We hope to examine these political and economic effects on their employees using data on 1,060 multinationals across the world and in employee data of the same multinational in Brazil. After receiving the panel data regarding the companies in Brazil, we have to perform data cleaning and extracts only the desired data to perform further analysis. This part is the main duty of the student assistant throughout the whole UROP1100Z. Data Collection and Cleaning for Topics in Health and Sports Economics Supervisor: LI Xuan / ECON Student: XU Runzhao / SSCI Course: UROP1100, Summer This project focus on data collection and cleaning for topics in health and sports economics on the following three aspects -- Performance, salary, contract, transfer data of soccer players from major professional leagues, results of National Games of the People's Republic of China and transactions and ratings data from https://www.haodf.com/. The whole project is still in progress and up to now, we have finished data collection and cleaning in the first aspect. The data being worked on will contribute to our comprehension of team cooperation and team performance. Data Collection and Cleaning for Topics in Labor Economics Supervisor: LI Xuan / ECON Student: TAN Yadan / ECOF Course: UROP1100, Spring UROP2100, Summer The whole UROP2100 course in Summer 2020 continues the ongoing projects in the UROP1100 course last semester, which included two parts -- Data Collection and Cleaning for Topics in Labor Economics, and Peer Effects: Evidence from Chinese Schools. For the former, I continued the search for the headquarter locations of multinationals with establishments in Brazil according to different legal natures. With all the wage fluctuation data available, this will help to identify whether macro environment changes in headquarters’ countries are one of the main causes of the wage fluctuations in Brazilian establishments, thus, to learn more about labor economics. For the latter, I translated the whole set of Chinese questionnaires (4 in total) into English for future modelling and wrote a Stata do file which could automatically convert all the collected data from Chinese into English version, so that facilitated subsequent data processing. School of Business & Management Department of Economics 121 Data Collection and Cleaning for Topics in Labor Economics Supervisor: LI Xuan / ECON Student: YE Muyang / ECOF Course: UROP1100, Spring UROP2100, Summer This is a progress report of the UROP project “Data collection and cleaning for topics in labor economics” supervised by Prof. LI Xuan. The purpose of this project is to complete preliminary data cleaning and operations for future analysis and research in labor economics related areas, like household and individual characteristics and their relation to people’s labor market participation and outcomes. I helped to conduct operations on census (obtained from IPUMS) and household survey (PNAD) datasets in Brazil. Additionally, I assisted with industrial code crosswalk. Peer Effects: Evidence from Chinese Schools Supervisor: LI Xuan / ECON Student: LI Jiting / MAEC Course: UROP1100, Fall Help the participant to acquire the personal experience as well as gross cognition on the complete process of applied economic research, in terms of generating and communicating the sentiment about the economic effects reflecting from the policy outcomes, questionnaire design and consummating, data collection and processing, delivering research thought and questions. Investigate the intuitive changes on account of the measurement been taken out on the Chinese schools, from the aspects extend to the corresponding involved children, teachers and parents to analyze the scholastic peer effects’ economic impact and value from the results embodied on The Sunshine Policy. Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: BIAN Ce / MAEC HE Jiaming / MAEC OU Shiying / ECOF Course: UROP1100, Summer UROP1000, Summer UROP1000, Summer This paper investigates the impact of BRI on China’s international trade with all countries and OECD countries as well as trade between OECD countries. Firstly, it analyzes the trade gravity effects among China, OECD countries and BRI member countries. Secondly, it explores the incremental effects on trade after the countries joined BRI using difference-in-differences model. It concludes that BRI significantly decreases the geographical trade barrier and enhances the pulling role of economic size on bilateral trade volume among China and BRI member countries. Import of China from BRI members and import between two BRI members among OECD countries increase since they joined BRI. School of Business & Management Department of Economics 122 Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: CHAN Chak Him / FINA GAO Zimo / ECON TANG Yunyang / MAEC TSANG Pak Ho / ECOF Course: UROP1100, Spring UROP1100, Spring UROP3100, Spring UROP1100, Spring COVID-19 pandemic is one of the most severe pandemics in 21st century. This disease causes around 400,000 casualties and most of the economic systems in the world are forced to shut down to avoid the spread of the disease. The shutdown disrupted the supply chain and international trade activities. This report provides a thorough econometric analysis on the amount of goods traded between China and other major economies and compare the amount with other period to investigate how the pandemic affect trading activities. The authors collected raw data of Chinese import and export value from the official website of General Administration of Customs of the People’s Republic of China, cleaned and analysed the data to find the dominant trading partners, dominant goods imported and exported, and the corresponding changes in amount of goods imported and exported. Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: GAO Zimo / ECOF ZHANG Ziyang / MAEC Course: UROP2100, Summer UROP2100, Summer The United States has implemented a set of tariff policies against China since 2018, comprising of 4 lists released in July 2018, August 2018, October 2018, May 2019, September 2019, December 2019. The authors studied the effect of the US tariff policy on China’s export to the US as well as its diversion effect on China’s export to third countries. The authors collected tariff data from 2015 to 2019, and constructed 3 groups of diff-in-diff models to study the effects of tariff policies. Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: HAN Chih-kang / ECOF YAU Chin Fung / MAEC Course: UROP1000, Summer UROP1100, Summer In a globalized world, different segments along a product chain cooperate with each other closer than ever. Therefore, the administrative sanctions on Chinese downstream companies are lethal to their production. We try to analyze the effect of the trading restriction announced by the US President upon the high-tech-related import of China. With the custom data set from 2010 to 2020 provided by prof Li, we are able to distinguish the growth and distribution quarterly, along with the comparison of the events timeline in real time. We will use graphs to indicate the trend and prove our reasoning about the effect. School of Business & Management Department of Economics 123 Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: LI Nga Yung / ECOF Course: UROP1100, Fall This progress report summarizes the proceedings of my UROP1100 project: Firms in Globalization: Evidence from China, in Fall Semester 2019. To analyse the level of globalization and transaction details of firms in China, I mainly focus on three types of database, the firm-level CEES file, the Chinese Industrial Enterprise data (CIE) and custom database. My main task is to create two combined dataset: the CIE dataset merging with CEES file and the Custom dataset merging with CEES file, and then summarize the overview of the matching outcome of these two merged files. The task aims to enable me to be familiar with the skills of modification and matching database through Stata. Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: LU Lingfei / ECOF ZHANG Shiyu / MAEC Course: UROP4100, Spring UROP1100, Spring In this project, we tried to identify the spillover of foreign direct investment (FDI) and examined its effect on the performance of domestic firms in China. At this stage, our focus is primarily on horizontal spillover from multinational firms to their domestic counterparts in the same industry. To measure the presence of FDI in an industry, we use output-weighted average foreign equity participation as a proxy. Our econometric analysis based on the firm-level Chinese Industrial Enterprise (CIE) data provides evidence in support of mixed effects of horizontal spillover. This report would serve as a presentation of the progress we made so far and guidance for interested readers to replicate and verify our results. Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: SHENG Ruoheng / MAEC WANG Shuoran / MAEC XIAO Zhihao / MAEC Course: UROP2100, Summer UROP2100, Summer UROP1100, Summer This paper empirically studied the trade pattern between Mainland China and Taiwan and the impact of a company’s relationship with its parent company on the company’s export ability. Through analysis on three datasets: Taiwan Invested Firms data (TIF data, hereafter), Chinese Industrial Enterprises data (CIE data, hereafter) and transaction data from the Chinese custom (custom data, hereafter), we discovered that Taiwan invested firms in Mainland China generally had more imports from Taiwan than exports to Taiwan, potentially indicating their reliance on their parent companies for sourcing. Meanwhile, a company’s closeness with its parent company seems to have a negative impact on their export ability. The result is to be verified with data of a longer time period in the future. School of Business & Management Department of Economics 124 Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: TANG Yunyang / MAEC ZHANG Ziyang / MAEC Course: UROP2100, Fall UROP1100, Fall US-China Trade War is an ongoing economic conflict between the world’s two largest national economics in the past years. And the conflict is highly displayed in two nations’ tariff policies. This report provides a thorough econometric analysis on US-China Trade War, from background and specific tariff policies stated by US and China to the impact on trade balance and further concerns. The authors collected raw data of Chinese import and export value (2015-2019) from the official website of General Administration of Customs of the People's Republic of China, did data cleaning, generated useful variables, and then did empirical data analysis and graphics using STATA (Version 15.1). Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: WANG Yimiao / MAEC Course: UROP2100, Fall Our work involves quantifying the spillover effects on domestic industries in China caused by foreign capital, because the knowledge and productivity improvement bought by the multinationals is always at the interests of local competitors, suppliers, or consumers. The spillover effects can be broken down into three types: horizontal spillovers between foreign and domestic affiliates, backward spillovers between foreign consumers and local suppliers, and forward spillovers between foreign suppliers and domestic consumers. Our study focuses on the inter-industry relationships caused by the inward foreign direct investment (FDI), using the firm-level Chinese Industrial Enterprise (CIE) data. The data indicates that, generally, FDI has non-negative horizontal effects on domestic competitors within the same industry. Except for food processing industry in 2002, foreign equity caused negative horizontal effects on domestic counterparts. Firms in Globalization: Evidence from China Supervisor: LI Yao / ECON Student: XIE Meng / MAEC Course: UROP2100, Fall This report is a progress report of a project in economics. The project aims to explore the economic behavior of manufacturing firms in Fujian Province based on three sets of data, network, CIE, and trade_yearly_name. The project focuses on the pattern of trade manifested in the transactions between firms in Fujian Province. The process involves merging three sets of data together and identifying patterns through the merged dataset. The statistics obtained include matching rate between different files, distribution of number of firms with respect to number of transaction partners, ratio of firms with mutual supply and demand relationship and ratios of transactions between domestic and foreign firms. School of Business & Management Department of Economics 125 The Economics of Car Sharing Supervisor: LIN Yatang / ECON Student: CHAN Cheuk Yin Rudolph / ECON Course: UROP1100, Fall In the big data era, developing advanced techniques in data collection could save significant amount of time in repeated work and provide more insights in the hidden pattern for the research project involved. The report discusses the current data collection work, dataset involved, potential challenges, and solutions. It will also recommend some advanced automatic programs built on python for the future application in the analysis of the impact of the proliferation of Uber and Lyft on job transformation in US. The Economics of Car Sharing Supervisor: LIN Yatang / ECON Student: SUN Hanqi / ECOF Course: UROP1100, Fall Nowadays, when the public enjoy the convenience the for-hire transportation service is bringing, cases that passengers are hurt by drivers also often happened. Some have come forward that no company uses fingerprints or law enforcement to background-check onboarding drivers is dangerous. Recently, American Congress passed a law that drivers with criminal records can’t be employed to decrease the probability of passengers getting hurt from drivers. While others think that if people with criminal records can’t even find the job as a driver, it is easier for them to offend again. Based on the data collected from incidents involving Uber and Lyft, this research aims to explore if this law had a positive effect to the whole society. The Economics of Car Sharing Supervisor: LIN Yatang / ECON Student: YANG Haohong / ECOF Course: UROP1100, Fall The rise and development of transportation network companies in the past few years seemed to bring some potential safety hazards. In response, most district courts and governments in the United States helped to enact the relevant laws and regulations. However, the effectiveness of these laws was still under discussion. Thus, the main task of this research program is to study the social and economic impact of car-sharing such as Uber and Lyft, in which the focus is to figure out the specific effect of the passage of driver background check laws on the crime rate related to online car-hailing by implementing data collection, integration and analysis. Then, applying the methodology of econometrics, we can test the hypothesis that whether the effect of the passage of the law on reducing the crime rate is explicit. The test results can help us answer meaningful economics and policy questions and offer suggestions to the government when it considers what will come next. School of Business & Management Department of Economics 126 The Environmental Cost of Power Shortage Supervisor: LIN Yatang / ECON Student: MASLIN Jason / ECOF Course: UROP1100, Fall This project is based on understanding the environmental cost of power shortages. The main tasks conducted was data collection and data cleaning. The environmental cost is estimated by gas pollution data given by 8 parameters (PM10, PM2.5, NO, NOx, NO2, SO2, CO, Ozone) while power shortages are given by the occurrence of outages during a day. The project focuses on data observed from the Delhi region of India for the period of June 1, 2019 to August 31, 2019, and were obtained from the Central Pollution Control Board as well as outage reports from three electricity suppliers in India (BRPL, BYPL, Tata). The data collected will be used to quantify the effects of outages on gas levels in the area. Social Networks and Financial Choices of Filipino Domestic Workers in Hong Kong Supervisor: VISARIA Sujata / ECON Student: IP Wang Hei / MAEC Course: UROP2100, Fall UROP3100, Summer This summer, we summarized our pioneer work last year. We tried to build up a research plan to see why the Filipino domestic workers in Hong Kong would choose to borrow instead of dissave for foreseeable expenditure. We tried to design a field experiment for this. To do so, we looked into the procedure of coming to Hong Kong to be a Overseas Filipino Worker (OFW). We also looked into some related literature to design intervention for the study. We also talked to some related persons to find out more information about the procedure of application to be an OFW and also employment agencies in both Hong Kong and the Philippines. Social Networks and Financial Choices of Filipino Domestic Workers in Hong Kong Supervisor: VISARIA Sujata / ECON Student: LEE Andrew Christian Tan / ECOF Course: UROP1100, Fall This paper reports the progress in determining why Filipino domestic workers in Hong Kong co-hold. The study’s rationale is to understand their financial choices and deduce their decision-making processes to offer concrete advice for their benefit in alignment with their personal goals. An experiment was conducted at HKUST to examine this, and the data collected was reviewed and analyzed further. After reading through relevant literature, it was confirmed that their social networks were instrumental, revealing the impact of kin taxes. However, as it could not explain the survey’s results entirely, the idea of reciprocity emerged from journals, personal interviews, and studies on their cultural backgrounds, which are being explored by designing reciprocity and expenditure modules to be administered next semester. School of Business & Management Department of Economics 127 Social Networks and Financial Choices of Filipino Domestic Workers in Hong Kong Supervisor: VISARIA Sujata / ECON Student: TANG Tsz Yeung Alvis / ECOF Course: UROP2100, Fall UROP3100, Summer In Summer 2020, our team planned for a linked panel data research to investigate the causes behind the indebtedness of Filipino Domestic Workers (OFWs). In this paper, I outlined multiple issues that arose in the planning stage, namely the sampling strategy, the intersection point, and possible interventions. Upon analyzing data from a Hong Kong employment agency, I identified Isabela, National Capital Region, Davao del Norte, and South Cotabato as possible main sending regions. 4 possible interventions were suggested, namely i) tied shopping, ii) repay and save and iii) financial literacy treatment, and iv) remittance control. Last but not least, I outlined the reported impacts of COVID- 19 on the Hong Kong Foreign Domestic Worker market. Social Networks and Financial Choices of Filipino Domestic Workers in Hong Kong Supervisor: VISARIA Sujata / ECON Student: TSOI Wai Jing / MAEC Course: UROP1100, Fall UROP2100, Summer Oversea Philippine Domestic Workers (OFWs) were observed that when they can save for their foreseeable expenditure in future, they choose to borrow; and they borrow even when they do have financial ability to pay. These phenomena are unusual and irrational because borrowing is more expensive than saving in that purchasing things through borrowing loan, one must pay extra money, interests. To determine the explanations, we designed an experiment, including survey and games, and interviewed 62 OFWs in last summer. In following fall semester, we did further data analysis and discussed about bettering the design of future experiment. This report will give a summary about the progress in current summer semester, and a thought about intervention of saving in the future experiment. Social Networks and Financial Choices of Filipino Domestic Workers in Hong Kong Supervisor: VISARIA Sujata / ECON Student: UY Rayner E mmanuelle Tang / ECON Course: UROP1100, Summer Overseas Filipino Domestic Workers are facing trouble in terms of saving money which leads to taking up debt in times when money is needed. So far, we have found that the process these domestic workers go through before moving to Hong Kong is more painstakingly convoluted than it should actually be and the seminars they received both in the Philippines and Hong Kong did not do much to increase their level of savings. The conditions for these workers have improved leaps and bounds than what is was 5 years ago but there are many more areas where their lives could be better. The project gathered quite sufficient data and information regarding their employment process and moving forward, a potential topic of discussion would be the interventions that would be administered in the actual research. School of Business & Management Department of Economics 128 Social Networks and Financial Choices of Filipino Domestic Workers in Hong Kong Supervisor: VISARIA Sujata / ECON Student: ZHANG Shiyu / MAEC Course: UROP2100, Fall Our project this semester is a continuation of previous work. Building on the analysis of previously collected data, we proposed and discussed hypotheses to explain our experiment outcome, more specifically, the difference of Octopus card consumption level among the experiment arms. Reflecting on our filed experience and relevant literature, we designed new survey modules for future fieldwork. This author designed a module to measure expenditure data, it contains retrospective questions for expenditure level before experiment and diary questions to record expenditure after experiment. The survey scripts and instruments are attached to the online Appendix, justification and interpretation of the design are discussed in this report. School of Business & Management Department of Finance 129 Department of Finance Corporate Risk Management Practices: Global Survey of BRICS-Country Firms Supervisor: MACKAY Peter / FINA Student: LAM Ka Chun / ECOF YU Shu-Han / GBUS ZHOU Yuchen / ECOF Course: UROP1100, Fall UROP1100, Fall UROP2100, Fall This project aims to investigate how BRICS-country firms (Brazil, Russia, India, China, South Africa) practice corporate risk management. A global survey research in seven languages (English, Spanish, French, Portuguese, German, Russian and Chinese) is conducted. The survey first identifies the different types of risks that a firm could be exposed to such as exchange rates, interest rates and financial derivatives and subsequently evaluates the impact of the risks on the profitability and financial health of the firms. The survey also investigates what kind of risk management measures are implemented by the firms. The project is divided into four major stages: survey planning, survey execution, data analysis and dissemination of results. Currently, the project at the beginning of the data analysis stage where all the data from the surveying stage has been collected and has undergone initial sortation and processing. The team has also translated some supplementary materials from English to Chinese. Corporate Risk Management Practices: Global Survey of BRICS-Country Firms Supervisor: MACKAY Peter / FINA Student: MIRZA Tinaz Firoze / ECON WANG Matthew Sebastian / ACCT Course: UROP1000, Summer UROP1000, Summer Corporations take different measures to manage their financial risk based on the stakeholders’ needs. The goal of this project is to analyze the impact of these strategies. Currently, the survey stage has already been executed, with results already having been received; the survey results will be analyzed with respect to the daily adjusted stock price to evaluate the impact of the risk management decisions made. The stock prices will be measured against a benchmark of the national stock index, bond prices, and the price of precious metals to isolate the impact of the firms’ financial decisions. This project is now in its final stages of data collection; the next steps will involve data pre-processing. Corporate Risk Management Practices: Global Survey of BRICS-Country Firms Supervisor: MACKAY Peter / FINA Student: ZHOU Yuchen / ECOF Course: UROP3100, Spring This project aims to investigate how BRICS-country firms (Brazil, Russia, India, China, South Africa) practice corporate risk management. A global survey research in seven languages (English, Spanish, French, Portuguese, German, Russian and Chinese) is conducted. The survey first identifies the different types of risks that a firm could be exposed to such as exchange rates, interest rates and financial derivatives and subsequently evaluates the impact of the risks on the profitability and financial health of the firms. The survey also investigates what kind of risk management measures are implemented by the firms. The project is divided into four major stages: survey planning, survey execution, data analysis and dissemination of results. Currently, the project at the beginning of the data analysis stage where all the data from the surveying stage has been collected and has undergone initial sortation and processing. The team has also translated some supplementary materials from English to Chinese. As the only research assistant at this data analysis stage, I mainly deal with some literature review work and help organize them. School of Business & Management Department of Finance 130 Generating Unbiased Financial News Articles Using Artificial Intelligence Supervisor: MACKAY Peter / FINA Co-supervisor: SAK Halis / FINA Student: CAI Mingqing / MATH-SFM Course: UROP1000, Summer Corporations take different measures to manage their financial risk based on the stakeholders’ needs. The goal of this project is to analyze the impact of these strategies. Currently, the survey stage has already been executed, with results already having been received; the survey results will be analyzed with respect to the daily adjusted stock price to evaluate the impact of the risk management decisions made. The stock prices will be measured against a benchmark of the national stock index, bond prices, and the price of precious metals to isolate the impact of the firms’ financial decisions. This project is now in its final stages of data collection; the next steps will involve data pre-processing. Risk-Management Duel: Financial vs Operational Risk Management Supervisor: MACKAY Peter / FINA Student: CHEN Kuan-chi / IS LI Mengyi / QFIN LI Shuhong / QFIN LIAO Caixing / ECOF Course: UROP1100, Fall UROP1100, Fall UROP1100, Fall UROP1100, Fall Hedging is a risk management strategy widely used by companies and investors to limit or offset probability of loss from fluctuations in the prices of securities, commodities, currencies and potential financial market changes. There are many types of hedging strategies that commonly be constructed via forward contracts, swaps, options and so on. There is no doubt that hedging plays a great role in corporate risk management, but many questions and uncertainties are still related to this topic. Therefore, our project is to uncover how companies hedge and what the outcome from their hedging strategy by analyzing the annual reports of oil refiners. The rationale behind is that the footprint of cashflow hedge accounting is recorded within the public information, which helps to solve the problem of the lack of data. Risk-Management Duel: Financial vs Operational Risk Management Supervisor: MACKAY Peter / FINA Student: LAW Tsz Hin / FINA LEE Sze Chun Vanessa / FINA LEUNG Ka Hei / FINA YOU Jifan / SBM Course: UROP1000, Summer UROP1000, Summer UROP1000, Summer UROP1000, Summer In normal company operation, it is common for companies to observe their financial performance to make diligent development decisions. Companies should always look for expansion opportunities of down-scaling options based on its financial profitability. In this project, financial and operational management data of gold mining companies over the past 20-30 years will be studied to identify the risk management relationship between these two areas. Now, this project is still in its exploratory stage and now only data relevant to the companies’ operation for example costs and production is collected and stored in a database. School of Business & Management Department of Finance 131 Analyzing CEO Compensation Contracts Supervisor: ZALDOKAS Alminas / FINA Student: CHUNG Wan Tong / FINA MA Jiexi / GBUS SHI Kaili / MAEC WANG Yuxi / QFIN XU Anqi / FINA Course: UROP1100, Spring UROP1100, Spring UROP1100, Spring UROP1100, Spring UROP1100, Spring The UROP project analyses CEO compensation contracts and explains how differences in these contracts might lead to some firms being more successful than others. The first part of the report explains the association between CEO compensation arrangements and collusive cases, identifying that a poorly designed CEO compensation contract may lead to collusion which is not only illegal but also reduces social welfare. The second part analyses the distribution of collusive cases and reaches several conclusions, in the aspect of industry distribution of convicted cases, regional vs. national/international cases tendency, individual vs. company defendant analysis, yearly distribution and trends of convicted cases, and defendant position analysis. Analyzing CEO Compensation Contracts Supervisor: ZALDOKAS Alminas / FINA Student: LEE Andrew Christian Tan / ECOF Course: UROP1100, Spring UROP2100, Summer Firm directors are very influential for they impact both short- and long-term business decisions. Besides power, these individuals are financially supported and reputationally benefitted by this profession. As such, engaging in the boards of multiple companies emerged as a viable career path. However, doing so in the same industry may carry anticompetitive implications, which may be detrimental to consumers. The project aims to delve deeper into horizontal directors and tacit collusion, including finding means to quantify them and the factors that incentivize their presence. Related research papers, news articles, and journals were read and analyzed. This paper reviews the project’s progress with discussions on conglomerates, firm centralization, and industry growth with extensions to COVID-19, interstate and international collusion. Analyzing CEO Compensation Contracts Supervisor: ZALDOKAS Alminas / FINA Student: LI Mengyi / QFIN Course: UROP1100, Spring Corporate governance is one of the most important factors determining the company’s success. Poor corporate governance practices including bad executive compensation packages may cause severe consequences like mismatches of interests among multiple groups of shareholders. Some criteria determining the bonus of the executive officers may lead to illegal activities like bid-rigging and price-fixing. The study will use data collected from publicly listed companies and past lawsuits to find out the hidden pattern behind the compensation packages and the intension to take part in illegal activities. School of Business & Management Department of Finance 132 Analyzing CEO Compensation Contracts Supervisor: ZALDOKAS Alminas / FINA Student: LO Tsz Ying / FINA Course: UROP1100, Spring This paper concludes our work during the semester, when we did data collection on CEO’s compensation contracts and number of convicted business in the United States from the year 2009 to 2019. As we are interested in the correlation between compensation contracts and convicted businesses, this report concludes several relationships between criminal and civil non-merger convicted cases, convicted businesses and economic growth, convicted businesses and compensation methods. Upon the end of the report, this paper justifies itself on why companies should give stocks to managers to avoid them participating in antitrust activities and harming the repetition of a company. Products, Supply Chains, and Finance Supervisor: ZALDOKAS Alminas / FINA Student: AGGARWAL Minal / FINA GOEL Kalpa / SENG LI Mengyi / QFIN Course: UROP1000, Summer UROP1000, Summer UROP2100, Summer This study primarily focuses on analyzing the impact of innovation across diverse fields. To break down this idea, the primary source of innovation used throughout this paper is new product introductions/launches. This is done through looking at product announcements of publicly traded firms in a variety of media/news articles and see how the stock prices of these firms react based on the public’s perception of the media reports. Extending this analysis further, the study derives some key observations. For example, new product innovation is correlated to higher competitive marketspace, larger firms within it, along with greater productivity. Lastly, the study observes that as innovation increases the number of people employed tends to decrease across industries. Products, Supply Chains, and Finance Supervisor: ZALDOKAS Alminas / FINA Student: LIN Yun-yu / FINA Course: UROP1000, Summer This report summarizes the progress and results of the UROP course during 2019-20 Summer. As our title suggests, our research focuses on the effect caused by closure of antitrust authority, DoJ offices, on firms’ tendency of collusion. We examine CEO’s compensation contracts, which can be linked with competitor’s performance in order to discourage competition, causing negative impacts on consumers. We further explore possible extension for the study, relationship between appointed horizontal directors and extent of competition and collusion, and capturing if two firms are competitors. Lastly, a summary composed by our team for the abovementioned study will be attached for reference. School of Business & Management Department of Finance 133 Products, Supply Chains, and Finance Supervisor: ZALDOKAS Alminas / FINA Student: ZHAO Xuehan / SBM Course: UROP1000, Summer There are a variety of indicators of innovation of a firm or an industry, such as the number of patents and forward citations of firms’ citations, R&D expenses and textual analysis of analyst reports on large firms. In our New Product project, however, we creatively adopt a new measure based on new product launches by public firms in the US. We found the link between new product announcements of firms classified by a convolutional neural network approach and the stock price reaction and further derived a few stylized facts about competition, labor market and productivity. During this summer term, our group first focused on studying the paper and doing some criticizing thinking. And then we carried out more literature searching to extract more related information. Furthermore, we brainstormed to come up with new algorithms how to extract target information, do more sentiment or textual analysis and track product announcements on the same product over time. Hong Kong IPO Market Supervisor: ZHANG Chu / FINA Student: XIAO Zejun / QFIN Course: UROP1100, Summer The project intends to research into the Hong Kong IPO market and to generate comprehensive knowledge of IPO market sentiment in the past 20 years. We will track 4 major factors: fundamentals, technical indicators, market environment, and psychological aspects. We want to provide a general picture of the overall Hong Kong IPO market. For the first period of the project, we mainly focus on collecting necessary data from HKEX, financial data provider, financial news platform and other searching engines. We also read published research papers on IPO related issues and obtain insights for our further research. From the fundamentals data we collected, we discovered some interesting patterns to be examined further in future. School of Business & Management Department of Information Systems, Business Statistics & Operations Management 134 Department of Information Systems, Business Statistics & Operations Management Algorithmic Trading Strategies Supervisor: CHEN Yanzhen / ISOM Student: CHAN Cheuk Nam / COMP LAM Sai Ho / FINA LUO Yuqing / QFIN SHEN Yue / IS YEUNG Hei Tsun / RMBI YU Jinghan / RMBI Course: UROP2100, Spring UROP1100, Spring UROP1100, Spring UROP1100, Spring UROP1100, Spring UROP1100, Spring In business, people usually rely on stock news to try to determine and predict the pattern of stock markets. However, the announcement released by the chief executives of the companies is not always reliable because those executives are highly likely to speak in two tongues to earn profits for their own. One way to detect whether the news is in alignment with the truth is to check the consistency of the executives’ payment. If the executive publishes the good news about the company and buys its stock, the news is reliable. If the executive publishes the good news about the company and sells its stock, the news turns out to be fake. In order to get informed of the pattern of the whole market, a large amount of the news and the stock exchange made by the corresponding executives should be matched and checked. This project is mainly about utilizing the data mining techniques to detect the stock patterns with the consistency of stock news and corresponding executives’ payment. Algorithmic Trading Strategies Supervisor: CHEN Yanzhen / ISOM Student: CHATTERJEE Kaushik / ECON LAI Yu Hin / ECOF LEE Jae Yeol / MATH-AM SIN Cheuk Ying / FINA ZHU Xinyu / QFIN Course: UROP1000, Summer UROP1000, Summer UROP1000, Summer UROP1000, Summer UROP2100, Summer In the hope to predict the movements in interest rate and option prices, this report highlights how the dynamic effects of Federal Open Market Committee (FOMC) announcements and press conferences effectively convey price-related information and contribute to fluctuation in asset prices and risk premiums. In this report, we gathered substantial amounts of datasets to pave ways for statistical analysis, summarized the findings from relevant research papers, and developed sophisticated tables outlining the correlations between the movement of interest rate, return of VIX, realized volatility to FOMC announcements. Acknowledging the valuable guidance from Professor Yanzhen Chen, this report marks a milestone providing solid grounds for running diversified sets of regression moving forward, contributing as a constitutional part of algorithm trading strategies. School of Business & Management Department of Information Systems, Business Statistics & Operations Management 135 Algorithmic Trading Strategies Supervisor: CHEN Yanzhen / ISOM Student: LAM Ka Kei / ECOF Course: UROP1100, Fall This Undergraduate Research Opportunity Program is the first half of a project that incorporates clues from press conferences into trading algorithms. In the first half, we focused on the Question & Answer session of the press conferences. For example, when the number of follow-up questions from a same journalist is high, this might indicate that the answer from the spokesperson, usually the president, is vague – implying uncertainties in the future or ineffectiveness of macroeconomic policies. Algorithmic Trading Strategies Supervisor: CHEN Yanzhen / ISOM Student: LAM Sai Ho / FINA LI Zhiying / RMBI TEH De Juan / IS WONG Hon Yu / OM YUN Anqiergerile Diana / WBB Course: UROP2100, Summer UROP1000, Summer UROP1000, Summer UROP1000, Summer UROP1000, Summer In our UROP Research, we are using data collection and regression analysis to determine whether small actions such as gestures, intonations from the CEO’s while delivering their speeches have a correlation with the price movements on various stocks and futures prices. Our Goal is to test whether voice features signals insider trading. Algorithmic Trading Strategies Supervisor: CHEN Yanzhen / ISOM Student: UY Matthew Johann Siy / QFIN Course: UROP1100, Fall Advances in machine learning and artificial intelligence have enabled the creation of increasingly complex and smart trading algorithms that use non-traditional forms of input data such as videos. The advantage of using video as opposed to more conventional forms of input data such as changes in the price of an asset over time is that it allows for the opportunity to capture qualitative and behavioral factors that wouldn’t be available through conventional input data. For example, by analyzing the recordings of the quarterly corporate earnings call, we can determine the emotions behind the CEO’s words to circumscribe the reliability and veracity of his words. Similarly, this project aims to use artificial intelligence to analyze the monthly Federal Open Market Committee (FOMC) press conference to discover potential new trading strategies. The Federal Open Market Committee sets the expectations for other participants in the market, by setting target ranges for the federal funds rate, to promote the national economic goals. Specifically, they are in charge of open market operations, which refers to the purchase and sales of securities in the open market by the central bank. Every month, the FOMC holds a press conference to give a summary of their findings and conclusions, as well as answer questions that the press may have regarding the markets in general. By analyzing the facial expressions of the Federal Reserve Chairman, we can gain insights into the emotions of the chairman as he answers the questions, allowing for a more comprehensive understanding of the answer. We can do this by tracking the eye movement of the federal chairman throughout the press conference. Further studies can explore the application of using this data to create new trading strategies that would need to be backtested. School of Business & Management Department of Information Systems, Business Statistics & Operations Management 136 Algorithmic Trading Strategies Supervisor: CHEN Yanzhen / ISOM Student: YIU Kwong Chun / ECOF Course: UROP1100, Fall In this research experience, the main part of my job is data collection. I have watched 15 videos of The Federal Open Market Committee Conference and I have to mark the details in the video, such as the location of the photographer, the name of the speaker, the reporters that raised the hand and the time when the speaker asked and answered the question. The above data could be used as the factor to decide whether the reporters that asked the questions are cooperating with the FOMC and making use of that can ensure there will be a less biased result. Algorithmic Trading Strategies Supervisor: CHEN Yanzhen / ISOM Student: ZHU Xinyu / QFIN Course: UROP1100, Fall This semester, I participated in the project of Algorithmic Trading Strategies. The project is aiming at making analysis on the videos of Federal Open Market Committee (FOMC) Conferences. I made marks for the videos from September 26, 2015 to January 30, 2019. The indicators include the date of the conference, the number of questions and answers, the names of the speakers, the institutes where speakers are working in, the beginning and ending time of questions and answers, the position of the questioners, the positions of all people that raise their hands, and the position of the camera. These indicators record the audio information for machines to learn and predict the adjustment of financial policies such as the rise or decrease of the interest rates. Apart from the analysis of audio information, there could be analysis of video information. Those indicators can be the directions of the sights, the movement of hands and so on. The video information is the second half of the project which I did not participate directly. Family Consent, Culture, and Organ Donation Supervisor: HUI Kai Lung / ISOM Student: DONG Yu / QFIN Course: UROP2100, Fall This report is a continuation of UROP1100. In this stage we first update the organ donation database as some countries change their laws regarding organ donation over time to boost the donation rate, then we focus on the statistical analysis using STATA, specifically the model-free analysis and fixed-effects regressions to understand how family veto, organ donation consent system, and culture affect the organ donation rate. Results show that the opt-out system boosts the organ donation rate and 4 out of 6 cultural indexes are statistically significant, the same as the law for family veto under certain circumstances. More data could be updated to the database and it’s worthwhile to study the interaction effect among the variables. School of Business & Management Department of Information Systems, Business Statistics & Operations Management 137 Dynamics in Crowd-Based Contests Supervisor: KOH Tat Koon / ISOM Student: CHUI Shuk Yiu / IS Course: UROP1000, Summer The research project focuses on the dynamic in the crowd-based contests. The crowd-based contests are often used by some companies to outsource the business project by providing the business requirement and reward for the platform users as incentives to join. The outcome of these contests often differs due to different aspects. This project will mainly focus on how the outcome of the crowd-based contests are affected by client-provided information and competition in the contest. The report consists of four parts - introduction about crowd-based contests, the mechanism of different crowd-based contests, different types of users using the platform, and the relationship between the competition and collaboration as the co-existed format of the crowd-based contests. Dynamics in Crowd-Based Contests Supervisor: KOH Tat Koon / ISOM Student: LIAO Caixing / ECOF Course: UROP1000, Summer Crowd-based contest is an emerging way of crowdsourcing in recent decades. Among different contest platforms, there are significant differences in submission quantity and quality, and winner selection. This research project aims to study factors that affect crowd-based contests outcomes. For example, we investigate how contest results are affected by client-provided information and competition in the contest. This report examines and categorizes various crowdsourcing contest platforms according to six dimensions. In the report, we also describe the various categories and how each of the platforms fits in in detail. As a result, we can see a clearer picture of the impacts of individual attributes on contest outcomes. A Hybrid Chatbot Model for Students Advising using Deep Learning Approach Supervisor: KWOK James Sai Ho / ISOM Student: CHEUNG Tsun Ming / ECOF Course: UROP1100, Fall In this project, we explored the foundation of building and implementing chatbot for student advising from scratch. The scope of the projects focused on handling course information enquiry and frequently asked questions (FAQ) in a domain specific context. We used Regular Expression (RegEx) to identify domain specific language such as course code, and NLP methods for question preprocessing. Pre-processed questions are then vectorized and compared with the database using Tf-idf. Lastly, after the matching, answers are retrieved from the pre-defined answer set corresponding to the question. School of Business & Management Department of Information Systems, Business Statistics & Operations Management 138 Deep Learning NLP in Finance Supervisor: YANG Yi / ISOM Student: LI Tianle / DSCT Course: UROP1100, Fall Nowadays, reviews and news from a variety of online shopping platforms and financial news media play a leading role of predicting the future trend of the market. And diverse corporations of large scale will take advantage of cuttingedge technology such as deep learning to make predictions. In the meanwhile, there are a good number of corresponding attacks to fool these systems in a way of decreasing the prediction accuracy of the deep learning model. In this research project, I exploited one of the attack methods which is generating adversarial examples to fool the Amazon reviews sentiment classification system. Deep Learning NLP in Finance Supervisor: YANG Yi / ISOM Student: WANG Shuoran / MAEC Course: UROP1100, Spring The rapid development of modern information technology has changed traditional finance industry a lot and Financial Technology (Fintech) is becoming an increasingly popular issue that has never failed to attract researchers’ and professionals’ attention. This project focuses on the applications of deep learning and natural language processing (NLP) techniques in finance industry and aims to implement the methodology on utilizing deep learning to predict stock price movement elaborated in the paper Stock Movement Prediction from Tweets and Historical Prices. The author self-learned deep learning and neural networks, conducted literature review on popular deep learning models for sentence classification and dealt with model setup based on the open-sourced model and dataset (https://github.com/yumoxu/stocknet-dataset). School of Business & Management Department of Marketing 139 Department of Marketing Luxury Marketing in Emerging Markets Supervisor: HELSEN Kristiaan / MARK Student: LING Xiuqi / ECOF ZENG Ziming / ECOF Course: UROP1100, Fall UROP1100, Fall Pop-up stores are temporary retail establishments that are designed to reach and engage customers through a creative experience. This report provides a thorough analysis on Pop-up stores in China, from the embedded marketing strategies and cost-benefit to real case analysis. In the marketing strategy, the authors extract the embedded strategies in pop-up stores marketing and show the effects. In the Cost-Benefit, the authors compare the cost of building a pop-up store to its various benefits. In the Case Study, the authors select four representative pop-up store cases for detailed Analysis. Then the authors summarize the positive and negative effects, reaching an overall evaluation of using such method of retail establishment and the possible trend of China market in the future. Luxury Marketing in Emerging Markets Supervisor: HELSEN Kristiaan / MARK Student: YOSHIDA Chieko / GCS Course: UROP2100, Fall The Chinese luxury market has been using social media marketing to reach target customers through key opinion leaders on different social media platforms. Following the research in summer 2019 with a qualitative approach, this research continues on the same topic from another perspective by analyzing the features of Chinese social media platforms and how brands utilize these platforms to achieve their marketing goals. To further enrich the study, two additional case analyses of Chinese KOLs are included and a pilot study on how KOLs influence Chinese consumers' buying decisions. Moreover, additional takeaways from the analysis will be provided at the end of the research paper. School of Business & Management Department of Management 140 Department of Management Multiculturalism, Social Judgments and Decisions Supervisor: CHAO Melody Man Chi / MGMT Student: GUNADI Vanessa / ACCT Course: UROP1000, Summer This report identifies the relationship between people’s culture, conflict management strategies, and Alternative Dispute Resolution (ADR). It suggests that people from individualistic and low power distance culture will have dominating conflict management strategies. They will put importance on distributive justice and prefer a third-party that can give them a coherence solution. This strategy leads them to prefer a binding ADR method. In contrast, obliging as a conflict management strategy will be prevalent to people from collectivism and high power distance culture. These people prefer a third-party that allow them to compromise and put greater weighting on procedural justice. Multiculturalism, Social Judgments and Decisions Supervisor: CHAO Melody Man Chi / MGMT Student: KIM Hayeon / CEGBM Course: UROP1000, Summer In globalizing world, international business interactions and formation of partnerships are increasing. Once the relationship forms, they are exposed to the possibility of facing conflicts. One way to resolve conflict well is to have ADR measure at the start of the relationships, so that when conflict actually arises, they are already in an agreement to how to resolve it. However, preferences for this method may differ due to many factors like power distance, perception of justice and third-party, identity of a disputing party, that influence the conflict management styles. Also, the degree of effectiveness of each factor may vary across different cultures and influence the opinion regarding ADR. If one does not know about the general perception towards ADR by the potential partner, implementation or even suggestion of ADR could backfire on them. Therefore, we aim to determine if there are any correlations between ADR preference and cultures by conducting survey in different regions. Social Perceptions, Justice, and Diversity Supervisor: CHAO Melody Man Chi / MGMT Student: IP Hoi Tung / SBM Course: UROP1000, Summer This paper focuses on understanding burnout and conflict. Burnout has become more prevalent nowadays, where it causes significant impact to individuals, their affiliated organization and non-work domains. Besides, increasing experience of conflict is also found to be a common phenomenon for individuals who suffered from burnout. Possessing a deeper understanding towards the association between burnout, conflict and conflict management style, would allow us to better control the impacts brought by burnout, achieve a better well-being on both personal and organizational level. This proposal explains the background, significance of conducting such a research, literature review and its correlation and extension to the proposed research. School of Business & Management Department of Management 141 Social Perceptions, Justice, and Diversity Supervisor: CHAO Melody Man Chi / MGMT Student: LEUNG Tsz Wan / BIBU Course: UROP1100, Fall Apart from the physical health, people nowadays focus more on their mental health and burnout is one of them. Burnout has been included in the organization’s International Classification of Diseases (ICD) and redefined by the World Health Organization (WHO) to be recognized as a form of work-induced stress in 2019, “a syndrome conceptualized as resulting from chronic workplace stress that has not been successfully managed”. Burnout starts to happen when people fail to cope with their stress. Not only will burnout affect the job performance, the physical and mental health, home, work and social life will also be affected. Therefore, we are reviewing different paper on the definition, antecedents, and consequence of burnout and find ways to predict it. Social Perceptions, Justice, and Diversity Supervisor: CHAO Melody Man Chi / MGMT Student: MARTINS Jessica Lynn / MGMT Course: UROP1000, Summer This paper will investigate whether anticipation of future conflict impacts an individual’s choice of conflict management strategy. The purpose is to determine if anticipating conflict, will affect how appropriately conflict scenarios are managed and thus resolved. Due to the lack of literature in this area, current research suggests there is uncertainty whether the anticipation of conflict produces a more positive or negative outcome on an individual’s health and well-being. The second part of this study will examine whether the anticipation of future conflicts impacts the occurrence of burnout. Literature has demonstrated a link between an individual’s experiences of burnout being influenced by their selection of conflict management strategy. Therefore, the objective is to study whether anticipation of conflict is an antecedent or alleviator of burnout. Social Perceptions, Justice, and Diversity Supervisor: CHAO Melody Man Chi / MGMT Student: PAN Xinyi / ECOF Course: UROP1000, Summer Generally, this report has an overall review on organizational justice literature. According to extant works, this report first introduces the mainstream model in justice literature – the four-dimensional model. This four-dimensional model greatly helps in building a clear and consistent system of organizational justice, and provides scholars a useful way to study on the consequences of justice. On the other hand, this report also talks about other three popular models and theories in justice literature (i.e. Fairness Heuristic Theory, Multifoci Theory and Uncertainty Management Theory), which emphasize more on “subjective” properties of organizational justice and may have more realistic applications. Finally, this report mentions how applying AI in organizations affects justice perceptions, which is the new trend of justice studies. School of Business & Management Department of Management 142 Responses to Globalization Supervisor: CHO Ja Ee / MGMT Student: YU Zherui / FINA Course: UROP1100, Summer Cultural appropriation happens when a person of one culture is taking something that belongs to people of another culture. This paper researches on people’s reactions to cultural appropriation. Articles from different US news media were reviewed and tested. Method includes data scraping, human coding, linguistic inquiry and word count (LIWC), and analysis of variance (Anova). The results showed that positive emotions were more expressed than the negative emotions. Also, it showed different levels of moral virtue vs. vice. Future studies could focus on the effect of the direction of cultural appropriation on people’s moral values based on these findings. Human Decision Making and Strategic Behavior in Social System Supervisor: DANIELS David Peter / MGMT Student: CHAN On Yu / IS ONGRATANAPHOL Ramon / MAEC YEUNG Chi Man / GBUS Course: UROP1100, Fall UROP1100, Fall UROP1100, Fall This report is written for the submission as final report of the UROP 1100 course in The Hong Kong University of Science and Technology. This report provides a review of our research findings and reflection during the research process. Prior research has extensively studied how human mind make decisions. Here, we examine different theories and studies, such as empathic accuracy, perspective-taking and the influence of emotion on interpersonal trust in social systems. The report tries to bring about a more precise and holistic view on the complex role of emotions in human decision making through conducting literature review. The information and data presented in this report has been collected from secondary sources. Data-Driven Sports Analytics Supervisor: KIM Yong / MGMT Student: LI Yaqiao / ECOF ZHOU Jianghua / ECOF Course: UROP1100, Spring UROP1100, Spring The objective of this study is to examine how injuries affect passing patterns and team performances in professional basketball. We quantify a team’s passing pattern of each game and compute its deviation from the season average. We find that, firstly, the number of injuries is positively correlated to the deviation from the season average passing pattern, and the deviation from the season average passing pattern is negatively correlated to the team performance. Secondly, the stability of lineup is positively correlated to the team performance. We also visualize passing patterns using an alternative approach, network graph. School of Business & Management Department of Management 143 Understanding Human Behaviour in the Workplace Supervisor: ONG Madeline / MGMT Student: CHENG Lai Him / IS Course: UROP1100, Spring The report is aspired to support or provide insights into Prof. ONG Madeline's current research, which studies punishment in the workplace. In the report, two major topics will be discussed. The first discussion will introduce a model to explain different levels of managers’ behavior toward punishment. Managers have three objectives in response to punishment: (1) protect the high-level managers’ interest, (2) shift the responsibility from corporate failure to low-level managers, (3) gain the credit from corporate achievements. The second discussion explains the concept of shared and common knowledge, the process of transiting shared knowledge to common knowledge in the workplace, and the pros and cons of public punishment. Understanding Human Behaviour in the Workplace Supervisor: ONG Madeline / MGMT Student: LIU Xinyi / SBM Course: UROP1100, Summer The below articles focus on topics related to implicit theories of morality. The first article introduced implicit theories and conceptions of morality. The second article discussed how implicit beliefs influence trust recovery. The third one disclosed belief about immutability of moral character and punitiveness toward criminal offenders. The last article analyze about implicit theories and issue characteristics as determinants of moral awareness and intentions. Understanding Human Behaviour in the Workplace Supervisor: ONG Madeline / MGMT Student: MANUPUTTY Josephine Angela / CENG Course: UROP1100, Spring This report aims to introduce the background of Conflict Minerals Law enforced by the Securities and Exchange Commission (SEC) to help minimize the conflict in Democratic Republic of Congo (DRC). Under this rule, companies are required to conduct due diligence to ensure that the tin, tungsten, tantalum and gold (3TG) minerals in their supply chain have not benefitted armed groups in DRC directly or indirectly. Furthermore, companies need to submit the official due diligence report to the SEC and publish the result on their website yearly. This research provides 12 criteria that have been developed to evaluate and rank companies’ compliance towards the law based on the report and the procedure to do so. School of Business & Management Department of Management 144 Understanding Human Behaviour in the Workplace Supervisor: ONG Madeline / MGMT Student: VYAS Anvita / MARK Course: UROP1000, Summer This report describes how different situations and incidents that threaten the organization’s identity and how the members of the organization deal with the threats. This report analyzes 4 articles that deal with organizational identity threat. The first summary talks about the reaction of members of the top 20 business schools on Business Week ranking and their way of selective categorization to deal with the dissonance to their organizational identity. The second summary presents the split identification that the members of the catholic church faced due to the threat to their organization. The third journal talks about the dissonance that the employees of British Petroleum faced during the oil spill crisis and how the members dealt with the threat to their organizational identity. The last summary talks about how members of an organization deal with negative media coverage and follow the organization threat response process to deal with the dissonance. Understanding Human Behaviour in the Workplace Supervisor: ONG Madeline / MGMT Student: XIA Xiaotong / MAEC Course: UROP1100, Spring In computer science, artificial intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and animals. Leading AI textbooks define the field as the study of 'intelligent agents': any device that perceives its environment and takes actions that maximize its chance of successfully achieving its goals. Colloquially, the term 'artificial intelligence' is often used to describe machines (or computers) that mimic 'cognitive' functions that humans associate with the human mind, such as 'learning' and 'problem solving'. In this report, I would like to talk about Artificial Intelligence and Ethical Concerns and elaborate more on how AI is being used in different domains in Mainland China (e.g., in gaming, healthcare, education, etc.). I would also expound some main ethical concerns and worries people have about AI in each domain based on the knowledge I collected. Understanding Human Behaviour in the Workplace Supervisor: ONG Madeline / MGMT Student: YEUNG Chi Man / GBUS Course: UROP1100, Summer Dan concluded that in this game, there is a good and bad equilibrium where there are no stable solutions in between. Yet the good equilibrium is weak in the sense that if anyone deviates or betrays the public then the whole equilibrium collapse. In the bad equilibrium, however, it is hard to restore the trust. Lastly, while government is injecting money, small companies may not get enough money and it is citizen’s role to keep them alive. Consumer and businesses are not in a competitive relationship but a collaborative one, where people’s lives are better off with businesses providing goods and service. Abstracts of UROP Projects 2019-20 SCHOOL OF HUMANITIES & SOCIAL SCIENCE School of Humanities & Social Science Division of Humanities 146 Division of Humanities Korean War Prisoners Oral History Interview Transcription Supervisor: CHANG David Cheng / HUMA Student: CHO Yeowon / GCS Course: UROP1000, Summer The overall objective of this project is to study Chinese soldiers’ and prisoners’ experiences in the Korean war. The main online archive I used is “The Archives of Korean History”, which provides various archival documents stored electronically. As a Korean, I have never thought of researching Korean War prisoners who were of different nationalities. I found this perspective very interesting. Korean War Prisoners Oral History Interview Transcription Supervisor: CHANG David Cheng / HUMA Student: CHU Chau Yuet / GCS Course: UROP2100, Fall Professor David C. Chang’s research project aims to record the oral history narrated by Chinese soldiers and POWs in the Korean War as well as Chinese interpreters in the Second World War. My duty is to transcribe the oral history interviews conducted by Prof. Chang. In the summer semester, I finished working on the history of Li Shiyu, a former Chinese People’s Volunteer soldier. In the fall semester, I continued to work on Zhang Buting, a former KMT soldier and a Korean War POW, and a newly interviewed WWII interpreter Feng Zhongliao, who is the son of the famous philosopher Feng Youlan. In the following report, I will examine the two interviewees’ oral history and my research experience. School of Humanities & Social Science Division of Humanities 147 Building Docent Tours on the Community Heritage of Yimtintsai Village, Saikung Supervisor: CHEUNG Siu Woo / HUMA Student: CHO Kwan Pui / GCS LEUNG Jessie / GCS Course: UROP1000, Summer UROP1000, Summer Yim Tin Tsai is an island in Sai Kung district and a village of the Chan Clan who belong to the Hakka dialect group. It was well known for outstanding heritage conservation, for example, the salt-pans received the Award of Distinction and the St. Joseph’s Chapel received the Award of Merit respectively in UNESCO Asia-Pacific Heritage Awards. Due to its uniqueness, it is significant to research on its Hakka culture and religious characteristics. This project mainly focused on one of the three major residential areas called ‘The Old Houses’ (老屋) on the island and found out its special characteristics, such as architectural significance of the deserted village houses and the Hakka traditional customs. Also, through interviewing villagers and conducting field trips, the research contains information about Hakka living styles and interesting past memories. Lastly, the reasons for the decline of Yim Tin Tsai is discussed as well. School of Humanities & Social Science Division of Humanities 148 Building Docent Tours on the Community Heritage of Yimtintsai Village, Saikung Supervisor: CHEUNG Siu Woo / HUMA Student: LAM Hoi Ching / GCS LOO Chung Wa / SBM Course: UROP1000, Summer UROP1000, Summer In the 20 century, as the economy develops, the Hakka culture in Hong Kong is getting less concerned. Yimtintsai village is a typical example where Catholicism was integrated with Hakka culture. However, all of the villagers had moved out at the end of the 1990s. In this report, we would place a spotlight back onto the culture. During the summertime i.e. June to August 2020, we have had both a face-to-face and an online interview with the former Hakka villagers on a field trip to Yim Tin Tsai to figure out their lifestyle, economic activities and customs. Information collected would be used to plan a Chinese docent tour to a residential area named ‘The New Houses’ (新屋) in Yim Tin Tsai. The route would be introduced with 4 topics of illustrations including their ways of entertainment, fishery, worship and agriculture in the corresponding spots. It was discovered that all of the above aspects have had immense changes. The last topic is the reflection after the interview with those former residents. School of Humanities & Social Science Division of Humanities 149 Building Docent Tours on the Community Heritage of Yimtintsai Village, Saikung Supervisor: CHEUNG Siu Woo / HUMA Student: WONG Sik Yu / GCS Course: UROP1000, Summer Yimtintsai, a Saikung village that was deserted by villagers in the late 1990s is revitalized in the 2000s. As a village that comprises of Hakka culture, religion (Catholicism) and the only remaining saltpan in Hong Kong, it appeals more and more visitors to the island especially after the conservation of St. John chapel and revitalization of salt pan successively won the “Award of Merit” and “Award of Distinction” from the UNESCO Asia-Pacific Heritage Awards in 2005 and 2015 respectively. The case of Yimtintsai cultural preservation generates profound implications for cultural preservation in Hong Kong and is worth studying. In the past summer, Professor Cheung recruited five students through the UROP program of HKUST to launch a heritage development project on village life in three former residential areas of the village community on the island, which had been neglected in the existing heritage tour program. I took the role as a research team member to interview the villagers in Yimtintsai, more specifically the villagers that had been lived in “The Little Alley” (巷仔) in order to be more familiar with the old times they spent there and collect information about village life in the past in a particular residential area of the community. The old lives of Yimtintsai in the past is embodied in this project and information is transcribed in favor of the docents who are going to conduct their heritage tours. School of Humanities & Social Science Division of Humanities 150 Eccentricity Behind Beauty: Psychopathology and Artistic Expression in Chinese Painting Supervisor: FU Li-tsui / HUMA Student: TONG Yvonne Chung Sze / GCS Course: UROP1100, Summer Over the past decades, the examination of psychopathology has evolved from an esoteric subject into a wellrecognised, even common, approach to understanding outstanding works of art. Regarded as one of the four Great Masters of the Yuan dynasty, the fourteenth-century Chinese literati painter Ni Zan (1301-1374) provides an excellent case study for exploring the intricate relationship between the psychiatric wellness of the artist, the development of his artistry, and the creation of his artworks. This research attempts to bring a deeper meaning to Ni Zan’s paintings as not only the ideal epitome of Chinese literati painting, but also the intriguing manifestation of the influence of psychosis on artistic expressiveness in a Chinese artistic and cultural context. With such goals in mind, this project proposes to investigate the underlying relationship between the presence of psychiatric disorders, or psychological eccentricity, and the growth of artistic creativity, against a synchronic narrative of landscape paintings created by Ni Zan in the late Yuan and early Ming periods. By situating Ni’s artwork and artistry in this new interpretative framework, this research seeks to shed different light on our understanding of Ni Zan as an artist through his actual psychological circumstances. School of Humanities & Social Science Division of Social Science 151 Division of Social Science GGS Survey Work: Localisation and Beta-Testing Supervisor: GIETEL-BASTEN Stuart Arthur / SOSC Student: LAI Man Ting / RMBI Course: UROP1000, Summer I am glad to participate and assist in the pilot study of the Generations and Gender Survey (GGS) in Hong Kong under Prof. Stuart Gietel-Basten. The GGS is an international survey instrument and is part of the Generations and Gender Programme coordinated by the Netherlands Interdisciplinary Demographic Institute (NIDI). It collects information on inter-generational relations, gender roles and work culture. As an undergraduate student, this is a very good training ground. Since this is the first time the GGS will be conducted in HK, I was given tasks focused on translations and localization. For the summer, I assisted in reviewing the translations, from English to Traditional Chinese. It covered translations of all questions as well as response categories. We also checked and ensured that all the survey contents fit the Hong Kong contexts. I was able to improve on my communication skills and language abilities by regularly interacting with other project team members who treated me as a colleague. My interpersonal and operational skills have also been enhanced by actively cooperating with them. Because of this rich and fruitful experience, I would like to continue to be involved with the conduct of the GGS-HK. Understanding Socioeconomic Determinants of Crimes in China Supervisor: HE Guojun / SOSC Student: CEN Yonglin / IS Course: UROP1100, Spring This report will show a student's three-month work on a socioeconomic research project 'Understanding Socioeconomic Determinants of Crimes in China'. There are three main tasks involved, Stata learning, research methodology learning, and the construction of a study on violence against doctors in China. In the Stata learning phase, the student has learned to use the tool for manipulating, analyzing and generating graphs for a data set. In the methodology learning phase, the student has looked into three methodologies in economic research studies. The three methods are randomized controlled trial, difference-in-differences, and regression discontinuity. Lastly, the student has shared her experiences in planning for a study and a brief overview of her future research plan. Understanding Socioeconomic Determinants of Crimes in China Supervisor: HE Guojun / SOSC Student: HONG Ka Ho / MAEC Course: UROP1000, Summer This is the final report for the research project – Understanding Socioeconomic Determinants of Crimes in China done by Max, Hong Ka Ho in 2020 Summer. This report will conduct a study mainly on the roles of the social assistance and social welfare on deterring the crime rates, I would kick off by investigating the root cause of crimes – Inequality by literature review and renowned theories. Then, I would head to the findings about how different types of social welfare scheme are contributing to reduce the crime rates and implications behind. Finally, I would extent the discussion to China by constructing a econometric analysis to see if social welfare scheme acts similarly in China. School of Humanities & Social Science Division of Social Science 152 Understanding Socioeconomic Determinants of Crimes in China Supervisor: HE Guojun / SOSC Student: SO Kwok Wai / ECOF Course: UROP1100, Spring This report provides an extensive literature review on the economics of crime. The first part of this review studies some of the socioeconomic factors that enter the decision-making process of potential crime commitment and criminal recidivism, namely education, Cognitive Behavioral Therapy and electronic monitoring. The second part focuses on existing economics theories on crime and camera surveillance, an increasingly used crime deterrent tool. Given the scale of China’s surveillance network being the largest in the world, its effectiveness on crime deterrence is worth investigating. Meanwhile, the judicial reform in China (2014-18) aims to enhance fairness, efficiency, standardization and transparency in the judicial system. Their interaction and joint effect on criminal activities could be dynamic. Understanding Socioeconomic Determinants of Crimes in China Supervisor: HE Guojun / SOSC Student: WANG Yilin / GCS Course: UROP2100, Fall UROP3100, Spring After learning by doing for several semesters, this semester is a period for command professional economic knowledge by understanding the mechanism of three useful econometrics methods, including randomized control trials, difference-in-differences and regression discontinuity. By self-learning with the materials provided by the professor, I am able to explain some application of these approaches in real studies. This report will give out a summary of 9 papers on how they adopt these methods. Moreover, to provide more insights into our research program, I went through the Judicial Reform Report from 2014 to 2017, and we select two topics for further data collection and investigation. Understanding the Costs and Benefits of Environmental Regulation Supervisor: HE Guojun / SOSC Student: YANG Mu / EVMT Course: UROP1000, Summer As a freshman of the project, I have studied five particular research papers related to environmental regulation and made a summary for each week. I learned how to conduct literature reviews and summarize statistical findings through reading these papers. Meanwhile, learning statistical software and data analysis is integral to the project, which is the foundation for conducting descriptive and exploratory research data collection. School of Humanities & Social Science Division of Social Science 153 Changing Alliances in Trumps White House: Networks and Elite Politics Supervisor: KELLER Franziska / SOSC Student: CHENG Yu / GCS Course: UROP1100, Spring This report is part of the “Changing Alliances in Trump’s White House: Networks and Elite Politics” research project conducted under the guidance of Professor Franziska Keller of SHSS. The goal of this UROP project is to try to automate the process of network analysis by teaching a computer how to recognize individuals and their relationship in the text, to achieve this, we have humans read enough sentences and determine the relationships described therein to create a dataset that is big enough to use it for machine learning, i.e. an algorithm to teach the computer to decide what relationships are contained in sentences in different texts. In this report, I will discuss my work in this project and the full report is divided into two parts according to my working stages: Biographical sentences collection and detection by the algorithmic website. The current U.S. administration has been facing unstable personnel arrangements since the beginning of the 2016 election campaign. Under the influence of Mueller's investigation and factional struggle, a large number of appointed officials under the president's office are in alliance or hostile situations: Frictions between the campaign team and the administration offices, changes in the relationship between the first daughter couple team and the chief strategist and several White House chiefs of staff are more and more exposed by biographies of writers and former officials. In order to better analyze the trend of personnel changes in the White House, this project collects materials from biographies, news, articles, and thus record the speech, behavior, and reflected sentiments between different people and institutions (named entities), so as to achieve the goal of network analysis with the help of machine learning. Next, I will discuss the operating principles of these two stages and how I contribute in the process. Changing Alliances in Trumps White House: Networks and Elite Politics Supervisor: KELLER Franziska / SOSC Student: PURI Sneha / ECON Course: UROP1000, Summer This progress report describes the progress of the research project that aims to identify the changing alliances in the Trump White House from the days leading up to the election and henceforth. The goal is to develop an algorithm that can correctly recognize and label these relationships through texts from a variety of sources. For this, a selected number of sentences were labelled manually and subsequently inputted in the developing algorithm to compare the two and identify areas of improvements for the latter. Overall, it was found that there were three main issues with the algorithm which needed to be improved. Changing Alliances in Trumps White House: Networks and Elite Politics Supervisor: KELLER Franziska / SOSC Student: YU Shu-Han / GBUS Course: UROP1100, Spring UROP2100, Summer This progress report describes the progress of a research project analyzing the network and relationships formed around the White House with Donald Trump’s Presidential campaign. The main focus of the project is to develop a coding scheme which can be applied to any written documents related to the networks around the White House and around the elite politics more generally. This progress report includes the progress of analyzing such relationships manually, building of datasets for relationship extraction that will ultimately be used to conduct social network analysis as well as the progress of categorizing and organizing the type relationships indicated. School of Humanities & Social Science Division of Social Science 154 Political Astroturfing in South Korea: Examining How the Secret Service Manipulated Twitter during the 2012 Presidential Election Supervisor: KELLER Franziska / SOSC Student: LEE Hojun / QSA Course: UROP3100, Fall UROP4100, Spring Keller, Schoch, Stier, and Yang identified temporal and message coordination patterns as unique characteristics of political astroturfing campaigns by analyzing Twitter contents of the National Intelligence Service (NIS) accounts during the presidential elections of 2012 in South Korea. We look for the coordination patterns in two electoral campaigns that were also conducted around this period and compare them with those of regular user accounts on Twitter. This paper provides a comprehensive overview of temporal and message coordination patterns of regular user accounts as a baseline to compare with those of political astroturfing campaigns and identifies a number of similarities and differences between the baseline and two other electoral campaigns. Political Astroturfing Worldwide: How to Manipulate Public Opinion on Twitter Supervisor: KELLER Franziska / SOSC Student: GAO Zhouyang / BCB Course: UROP1000, Summer Political astroturfing aims to manipulate public opinion. This report examines 795 twitter accounts and 2,700,296 tweets of an Indonesian information campaign on Twitter. Various general information about this dataset, including account and Tweet activities, most active language, hashtags, and users, were analyzed. The findings show that accounts in this dataset perform differently from regular users. Large number of tweets, high frequency of tweeting, organized hashtag usage all suggest that the accounts are from a well-organized campaign. An analysis of the hashtag activity reveals the campaign’s regulation on the format of tweets, number of daily tweets, and tweeting intervals. Political Astroturfing Worldwide: How to Manipulate Public Opinion on Twitter Supervisor: KELLER Franziska / SOSC Student: PANG Hong Man / MAEC Course: UROP1100, Fall UROP2100, Spring In this UROP Report, selected findings regarding analysis of the astroturfing campaign based in Hong Kong through the perspective of Hashtag usage is presented. It is found out that in the astroturfing campaign, there exists a type of tweets that are only a combination of hashtags and do not have their languages identified if they contained mainly English hashtags. A high proportion of Japanese tweets is also identified in this astroturfing campaign, with the discussion mainly led by key influencers that are on protestor’s side of the argument, a method to help improve accuracy of correctly identifying Japanese tweets in lieu of Twitter’s own algorithm is also proposed. School of Humanities & Social Science Division of Social Science 155 Political Astroturfing Worldwide: How to Manipulate Public Opinion on Twitter Supervisor: KELLER Franziska / SOSC Student: SHIN Mugyeong / QSA Course: UROP1000, Summer Aleksandar Vucic the Serbian Progressive Party (SNS) have been accused of using governmental authority to manipulate approval ratings and public opinion using political astroturfing. Previous research from the Stanford Internet Observatory identified the suspicious movement on twitter by analyzing the tweet dataset and provided evidences that Serbian government is related to the operation- something which it has denied. This report adds information from a whistleblower and interviews with a Serbian antigovernment journalist who was attacked by the suspicious movement. These interviews were conducted by BIRN (the Balkan Investigative Reporting Network). This report also analyzes suspicious patterns in the dataset provided by the Twitter to figure out the relation between Serbian government and the operation. The suspicious movements are arranged into different sort of charts and table. The report is the part of the UROP 1000 project, ‘Political Astroturfing Worldwide: How to Manipulate Public Opinion on Twitter’. I’ve chosen to examine the Serbian operation. I tried to analyze and converted the dataset, containing 43 million tweets and 8,500 accounts, into scatter plot and tables through R studio. I also included interview from a victim and whistleblower. The summaries of the interview and disclosure will demonstrate that the operations were not legitimate, and instead it was systematic. Political Astroturfing Worldwide: How to Manipulate Public Opinion on Twitter Supervisor: KELLER Franziska / SOSC Student: TANG Hoi Yan / GCS Course: UROP1100, Fall U ROP2100, Spring A Chinese disinformation campaign, aiming to sow discord about the 2019 Hong Kong Extradition Law Protest, was brought to light on August 19, when Twitter announced the suspension of 936 accounts involved in “coordinated inauthentic behaviour” (Chen, 2019). Although the Chinese government has been asserting that the aforementioned accounts were controlled by individual patriotic users, scholars investigating the accounts’ behaviours have suggested otherwise. Given that the disinformation campaign is a coordinated one, it is hypothesized that delayed discussion of topics might be observed among them compared to the normal users since they are likely to respond after receiving central orders, whereas individual users react right after the incident happened. In addition, discussion among astroturfers may be less sustained because of their simpler network of retweeting. This paper, therefore, will analysis on the differences in the posts appearing time and retweeting network among astroturfers and normal users to find out possible delays and the network. School of Humanities & Social Science Division of Social Science 156 Political Astroturfing Worldwide: How to Manipulate Public Opinion on Twitter Supervisor: KELLER Franziska / SOSC Student: WU Hongsen / GBUS Course: UROP1100, Spring Astroturfing campaigns are intentional attempts to inflate the level of public support on social media. Here we examine a method to detect and distinguish the astroturfing campaign on Twitter during Hong Kong protests in 2019. We applied a comparative approach to characterize behavior of astroturfing and normal accounts, based on the use of politics-related hashtags. The analyses recovered that astroturfing accounts do not frequently use politics-related hashtags, and their tweeting patterns have certain characteristics. On the contrary, normal accounts display different characteristics in using politics-related hashtags, varying across users’ residence countries. Such differences in political tweets thus provide insights about improving astroturfing content detection and understanding the mechanism of public opinion manipulation in astroturfing campaigns. State Visits and International Get-Togethers: Studying the Meetings of State Leaders Worldwide Supervisor: KELLER Franziska / SOSC Student: YEUNG Wing Ching Vinnci / IS Course: UROP1100, Summer Meetings between states leaders is always a big news to both countries involved, as signing of important trade or military agreements always happen during such meetings. The Project “State visits and international get-togethers: studying the meetings of state leaders worldwide” aims to find out the factors correlating or causing these meetings between leaders of different countries. In order to do so, a set of data with such visits need to be compiled. We are in the progress in compiling such list using a range of sources, including official websites of the foreign departments or head of the state’s office, third parties platforms as Wikipedia, Nexis Uni, search engine as Google and so on. In this progress report, findings and challenges during the progress will be discussed. The Spread of Rumors On and Across Social Media Platforms during Hong Kong's Anti-Extradition-Law-Protests Supervisor: KELLER Franziska / SOSC Student: TAM Sum Yee / GCS Course: UROP1100, Spring In Hong Kong’s year-long Anti-Extradition Law Amendment Bill movement, social media has played a vital role in terms of (i) spreading rumours, (ii) intensifying the conflicts between the two sides, as well as (iii) shaping both local and international public opinion. To gain an insight into the role of social media in the propagation of rumours and the factors affecting its speed, this research project covers topics ranging from (i) the major rumours being circulated across different platform to (ii) the preference of rumours that each platform users tend to spread. In this progress report, I describe the fact-checking sites mentioned in the project and their possible biases. School of Humanities & Social Science Division of Social Science 157 The Spread of Rumors on and across Social Media Platforms during Hong Kong's Anti-extradition-law-protests Supervisor: KELLER Franziska / SOSC Student: WU Hongsen / GBUS Course: UROP1000, Summer The Anti-Extradition Law Amendment Bill (Anti-ELAB) movement has caused social divisions and unrest in Hong Kong society, spreading of rumours on social media being one of the most vicious catalysts. During the spread of rumours, fact-check conducted by third party platforms is considered as one of the most important factors to either accelerate or hinder the spreading. The platform, methodology, timing and possible biases of such fact checking practices may influence their effects. This report searched the 27 rumours selected in the project on 8 fact-check websites. We firstly summarized how these rumours had been fact-checked, and then illustrated the variations across fact-check platforms with an example. We also analysed the elapsed time between the emergence and fact-check of rumours. Finally, we proposed possible suggestions for further research. The Socioeconomic Impacts of Environmental Issues in Hong Kong Supervisor: LIU Tong / SOSC Student: LEUNG Ho Long / MAEC Course: UROP1000, Summer Pre-existing studies have not reached consent on the effects of public transport on the housing market. Moreover, most of them only consider the direct effect of increased convenience but not the indirect health risk caused by air pollution from public transport (e.g. Tian et al. 2017). This study aims to understand both the capitalization effect of increasing bus accessibility and the effect of air pollution created by buses. The project is in progress, and the cleaning of bus data is completed. It is expected to combine existing data with air quality data and housing transaction data in the future to estimate the willingness to pay for bus accessibility and minimizing health risks. The study will provide a well-round analysis of the benefits and costs of public transportation. The Socioeconomic Impacts of Environmental Issues in Hong Kong Supervisor: LIU Tong / SOSC Student: WAN Cheuk Yin / EVMT Course: UROP1000, Summer The temporal and spatial patterns of COVID-19 cases in Hong Kong are briefly described in this work-in-progress report. To understand the possible factors of COVID-19 infection, the socioeconomic conditions of patients are also discussed. As this research is still under process, possible future study is also suggested. At last, my learning progress is also explained. School of Humanities & Social Science Division of Social Science 158 Belt and Road Initiative Supervisor: PARK Albert Francis / SOSC Co- supervisor: CHEUNG Chi Man / SOSC TRITTO Angela / IEMS Student: DU Yumeng / RMBI Course: UROP1100, Fall As a large volume of trade and investments under the Belt and Road Initiative goes, the efficiency and effectiveness of the program meet controversy with data from different sources. This report mainly reveals how the data were merged and consolidated to derive evidence-based insights on the impacts of OBOR program throughout the Undergraduate Research Opportunities Program. The report first shows the data collation process that contributes to the statistical analysis of OBOR investments and transactions. Then it discusses the sentiment analysis process which involves sentiment label tagging as well as data wrangling. At last, it summarizes the process and harvests of the research program. Belt and Road Initiative Supervisor: PARK Albert Francis / SOSC Co- supervisor: CHEUNG Chi Man / SOSC TRITTO Angela / IEMS Student: KIM Jungyoon / ECOF Course: UROP1100, Spring Belt and Road Initiative (BRI), being the biggest global development strategy that the Chinese government launched so far, is one of the most important agenda for the Chinese government. Notably, state-owned enterprises (SOEs) are taking a huge portion within the BRI, taking more than 70% of the combined value of contracts launched under the BRI, even though SOEs are generally perceived to be less efficient compared with the private-owned enterprises (POEs) in various dimensions such as capital allocation. This research suggests various reasons that could explain this phenomenon such as the degree of access to financial resources, resource-seeking motivations, overcapacity problem of SOEs, political motivations of the managers of SOEs, and a lack of alternative enterprises. Greater Bay Area Supervisor: PATCHELL Gerald Roe / SOSC Student: CHENG Tsz Yeung / ECOF Course: UROP1000, Summer Cultural Preservation and Urban Development has been a controversial topic in the Old City, Guangzhou since the Development Plan for the Guangdong-Hong Kong-Macao Greater Bay Area in 2017. Research has shown that many valuable historical heritages have been destroyed and neglected after the urbanization in Guangzhou. This study aims to analyze the diversity of historical heritages and the potential market value of culture commercialization. In this context, historical heritage indicates the buildings or streets with historic and artistic values. Cultural commercialization is defined as generating economic benefits from promoting the cultural value of the historical heritage. Based on the review on the Development Plan for the Guangdong-Hong Kong-Macao Greater Bay Area and literature reviews on cultural preservation in Guangzhou, analysis is conducted to illustrate the effectiveness in cultural revitalization and commercialization in Guangzhou. On this basis, it is recommended that local government use social media and technologies to promote the traditional Ling Nan cultures (嶺南文化). School of Humanities & Social Science Division of Social Science 159 China/Africa Links Project Supervisor: SAUTMAN Barry Victor / SOSC Student: MA Chin Long Natalie / SSCI Course: UROP1000, Summer This research focuses on one of Zambia’s emerging industries: agriculture. The development of this industry has simultaneously increased the country’s gross national income whilst bringing employment opportunities that can help alleviate poverty. However, Zambia has been facing physical challenges with lack of rainfall in the previous years which has led to downtick in agriculture production. Throughout the recent years, China and Africa has built a business-oriented alliance. Africa is an emerging market allowing China to benefit by growing its global presence whilst providing economic opportunities in return. Specifically, China saw potential in Zambia’s agricultural sector and has made several investments in this country such as the Solar Hammer Mill Initiative and the Scaling Solar Program. China/Africa Links Project Supervisor: SAUTMAN Barry Victor / SOSC Student: YUEN Man Chee / EVMT Course: UROP1100, Summer With the introduction of the Belt and Road Initiative (BRI), China Africa links has emerged to receive more attention in the international agenda. In particular, this researched is devoted to the Chinese involvement in Ethiopia development in the form of foreign direct investment (FDI). The research has examined the 3492 FDI projects for Ethiopia between 1992 – Jul 2020, looking into the nature and structures of the investment projects. Data provided by the Ethiopian Investment Commission (EIC) was analysed cross-sectionally and longitudinally to demonstrate the increasing Chinese presence since 2013, when BRI was formally introduced. The 'Yellow Peril' and Covid-19 Supervisor: SAUTMAN Barry Victor / SOSC Student: LIAO Dankai / MAEC Course: UROP1000, Summer The project is about the social aspects of the Covid-19 pandemic. Western countries, especially their politicians and media, have claimed that the spread of COVID-19 pandemic, was mostly caused by Chinese eating habits and failing in preventing the virus in last December, which actually has no factual basis. This project mainly examines detailed data about each of these notions and compares actions adopted by the Chinese government with other countries and people. School of Humanities & Social Science Division of Social Science 160 The 'Yellow Peril' and Covid-19 Supervisor: SAUTMAN Barry Victor / SOSC Student: LIU Ka Hei / SBM Course: UROP1000, Summer COVID-19 has hit the world suddenly and tremendously. Recently, Western leaders claim that China has spread COVID- 19 around the world. China has been criticized internationally. The purpose of this project is to examine these notions with empirical data. Students are required to use different academic search engines to collect data, such as ProQuest and Factiva. Journal articles and newspapers have been utilized to find out the origin of the first cases and most cases for every country and region around the world. Besides, we have analyzed and compared the precautionary measures taken by the US, UK and the Chinese government in response to the pandemic. We believe that most of the notions are false. The 'Yellow Peril' and Covid-19 Supervisor: SAUTMAN Barry Victor / SOSC Student: WANG Kaiwen / SSCI Course: UROP1100, Summer This project focus on social aspects of the COVID-19 pandemic. Ever since the massive outbreak, western bodies have been claiming that China is responsible for the spread of the disease throughout the world. We examine whether the above assertions agree with factual evidence by gathering data of the imported COVID-19 cases for each country and determining the role China had played in the spread of the virus. This project also covers some comparison of actions taken by Chinese and other countries’ government to evaluate the effectiveness of different measures with an attempt to elaborate the reason behind the infection rate and death toll. Citizen Science in the Post-truth Era Supervisor: WONG James K. / SOSC Student: CHAN Yin Tsit / GCS Course: UROP1100, Spring In view of the resemblance between SARS (2003) and COVID-19 (2019), Hong Kong with its traumatic experience with SARS had been expected to provide reference for other regions on counter-pandemic measures. Nonetheless, while Hong Kong demonstrated effective epidemic control targeting the outbreak of COVID-19, a general tendency of selective expert denial against government-related expertise had been observed among the Hong Kong society. This study elaborated on this phenomenon by comparing the discourse of governmental and nongovernmental expertise in Hong Kong and illustrating citizens’ response regarding counter epidemic measures from three aspects: border disclosure, use of mask, and social distancing. The conclusions implied that voluntary awareness and social cohesion were critical when countering major common threats in the absence of effective leadership. School of Humanities & Social Science Division of Social Science 161 Framing 'Democracy' in Hong Kong Supervisor: WONG James K. / SOSC Student: CHAN Hei / MGMT Course: UROP1000, Summer This study serves as a preliminary research to explore new perspectives in analysing the political development in Hong Kong, in view of the dynamic and expeditious transformation over the last few years. The political thinking of contemporary theorists Ernesto Laclau is introduced to lay the foundation of a dispute over the deterministic view in seeking a universal truth in politics. The review of the Laclau’s political theory is followed by an attempt apply his conception of ‘the political’ built upon antagonism and radicality to the political situation in Hong Kong, in which resonance is observed. An outline of the possible prospects for further study is presented as a conclusion. The Ethics and Governance of Emerging Technologies Supervisor: WONG James K. / SOSC Student: TANG Hoi Yan / GCS Course: UROP1100, Summer Different from traditional lampposts, multifunctional smart lampposts can be used for collecting traffic, weather and air quality data as well as providing Wi-Fi services. The HKD$270million-worth project of constructing 400 smart lampposts in 3 years, in fact, has been receiving criticisms since its debut. Meanwhile, the crucial point of contention is whether the smart lampposts would compromise privacy. Even with repeated clarifications from the government, doubts that the lampposts would be used as surveillance tools are never dispelled. This paper, therefore, will summarize 3 major arguments from the public challenging the building of smart lampposts and the responses from the supporters of the project. It will start with controversies over the cameras to be installed on the lampposts, followed by the RFID technology on them that could allow contactless extraction of personal information, and lastly, the supplier of Bluetooth beacons. The Ethics and Governance of Emerging Technologies Supervisor: WONG James K. / SOSC Student: TANG Kevin / ECOF Course: UROP1100, Summer The emergence of cryptocurrencies is far from an ephemeral fad, as vindicated by their substantial growth in the past decade. Their global outreaches and disruptive potentials have warranted considerable attention from experts, leading to heated discussions of their ethical applications. The highly controversial nature of cryptocurrencies also poses a huge conundrum for the government legislatures who have to decide whether cryptocurrency transactions should be regarded as basic rights enshrined in civil liberty, or whether they should be banned for their potential harms to the society. This report aims to present the handlings of cryptocurrencies in Mainland China (abbreviated as “China” hereafter) and Hong Kong to demonstrate the ethics and governance of cryptocurrencies. School of Humanities & Social Science Division of Social Science 162 Identifying Protest Events in China with Social Media Data Supervisor: ZHANG Han / SOSC Student: WANG Yiwei / QSA Course: UROP1100, Summer This UROP1100 report is part of the UROP project: Identifying Protest Events in China with Social Media Data conducted under the supervision of Professor Zhang Han, School of Humanities and Social Science. This project aims to study the patterns of protest events in mainland China, specifically how grievances become collective action. For the first stage of this research, Prof. Zhang has classified 9.5 million Weibo (the Chinese Twitter) posts from 2011 to 2017 using the CASM (collective action from social media) system. For the following research, more information related to every single event is needed to understand the process. This report mainly discusses the two different work stages I have done in this project: Weibo posts identification and the completion of related information for sequences from grievances to protest. Abstracts of UROP Projects 2019-20 INTERDISCIPLINARY PROGRAMS OFFICE Interdisciplinary Programs Office Division of Environment & Sustainability 164 Division of Environment & Sustainability Developing Computer-generated Images for Future Hong Kong Wind Farms Supervisor: DELINA Laurence / ENVR Student: LO Matthew Chi Chung / SENG Course: UROP1000, Summer Hong Kong’s own renewable energy future will likely rely much on offshore wind power. Compared to onshore wind turbines and other proven renewable energy sources such as solar and hydropower, the construction of offshore wind turbines has historically been prohibitively expensive, thus rendering them less attractive and riskier to invest into. This expensive cost is largely due to the challenges to build robust wind farms in waters deeper than 60 metres, which are the part of the ocean that tends to offer the greatest wind speeds, and where, Hong Kong’s future wind farms will most likely be built. Mr Matthew LO produced for this UROP project projections or images of these future offshore wind farms. These images are visions of how market-available wind turbines could look like if they were built in selected Hong Kong’s waters. Interrogating Fintech-Enabled Solar Power as Energy Democracy in Hong Kong Supervisor: DELINA Laurence / ENVR Student: LAM Rainbow Yi Hung / EVMT Course: UROP1000, Summer This UROP project reviewed the extant literature on blockchain based solar power market as a part of a longer study that seeks to uncover how energy democracy is produced in Hong Kong via a fintech-enabled renewable energy market. Ms Rainbow LAM reviewed 16 academic and grey articles and found the dearth of literature on Hong Kong’s use of fintech in the renewable solar market, suggesting that the technology is still nascent at HKSAR. The review suggested the potential and multiple benefits of fintech-enabled renewable energy market for Hong Kong, as well as the challenges that need to be hurdled to realize this potential. Air Quality Data Analysis and Interpretation Supervisor: GU Dasa / ENVR Student: WANG Yiyi / ECOF Course: UROP1100, Summer In recent years, our daily life has always been shrouded in rumours and fears about air pollution, but if investigating into related data, we can see that it’s not really that bad--the variation trend of pollutant concentrations isn’t always increasing--explore how their concentrations are affected by some common factors, and figure out best countermeasures for each type. Based on the air quality monitoring data of Hong Kong Environmental Protection Department and some basic statistical methods such as calculation of Pearson correlation coefficient, this paper briefly analyzes the variation patterns and influencing factors of the concentration of 7 kinds of monitored pollutants, and explains them by using some simple chemical principles. Interdisciplinary Programs Office Division of Environment & Sustainability 165 Air Quality Measurement Techniques Supervisor: GU Dasa / ENVR Student: LEUNG Cheuk Pui / CHEM Course: UROP1100, Fall This is a literature review of vehicular volatile organic compounds (VOCs) emissions in Hong Kong. Since vehicular emissions (VE) have been one of the major emission sources of VOCs in Hong Kong, several papers published from 2011 to 2016 were reviewed to study the vehicle VOCs contributions with respect to their fuel type. Carbonyl compounds emission was reviewed, seasonal patterns and diurnal trend were observed. Moreover, the contributions of VOCs to the ozone formation potential (OFP) were reviewed to highlight their important roles to the ambient ozone formation, in particular the liquified petroleum gas (LPG)-related VOCs contribution to OFP. The effect of substantial VOC profile change is briefly discussed as well. Modelling of Aquaponic System Supervisor: LU Zhongming / ENVR Co-supervisor: LING Li / CIVL Student: RENANINGTYAS Priyanka / SENG Course: UROP1000, Summer Aquaponics, a relatively new technology, is a system that integrates two better known technologies: aquaculture and hydroponics. It involves the practice of raising fish and using the effluent water from the fish tank to cultivate the crops. However, there is still a lack of knowledge about how to optimize aquaponic systems to ensure maximum efficiency. Therefore, the aim of this project is to develop models which can simulate the operations of an aquaponics system. The models established would then allow one to evaluate an existing physical system and study how altering various parameters would affect both the system’s economic and environmental efficiency and sustainability. The set parameters that induced desirable results could then be applied to improve the existing system. Use WiFi Technologies to Predict Building Energy Consumption Supervisor: LU Zhongming / ENVR Co-supervisor: QU Huamin / CSE Student: WANG Weiqi / COSC ZHOU Zixuan / EVMT Course: UROP1000, Summer UROP2100, Summer In the UROP project, our group has made improvements on previous work’s methodology and application. Please see the detail in the graph abstract below. Interdisciplinary Programs Office Division of Environment & Sustainability 166 Use WiFi Technologies to Predict Building Energy Consumption Supervisor: LU Zhongming / ENVR Co-supervisor: QU Huamin / CSE Student: ZHOU Zixuan / EVMT Course: UROP1100, Fall In the UROP program: Use WiFi technologies to predict energy consumption in the building, I mainly collected and composed the data for lateral programs, including both energy data (electricity) and climate data. Using the energy panel website developed by HKUST sustainability sector, I retrieved hourly electricity data in UG hall six as reference for prediction validation. I collected the climate data including temperature, relative humidity, irradiance and precipitation from the website of HKUST ENVF. For small number of data, I used Excel to compose the data and anonymized it. For large number of data, my groupmate helped me to use a program to read the raw data and composed them. Smart Campus Air Network (SCAN) Supervisor: NING Zhi / ENVR Student: CHUNG Shun Lap Benedict / ENVS YIP Wun Yu / EVMT Course: UROP1100, Fall UROP1100, Fall Through conducting literature review, an indoor air quality grading system covering CO2, PM2.5 and PM10 is suggested in this report. In the light of the lack of official indoor air quality grading system, standards and guidelines from various regions such as Asia, Europe, North America and worldwide are summarised, and hence used to interpret a reasonable grading system that may be applicable to the Smart Campus Air Network (SCAN) project in the future. This report aims at giving the future users of the SCAN a qualitative idea of the campus indoor air quality. Further research on the exact correlation of indoor air contaminants and the occupants’ health can be focused in the future. Ultra Compact Air Sensor Development for Personal Health and Indoor Air Quality Supervisor: NING Zhi / ENVR Co-supervisor: LAU Alexis Kai Hon / ENVR & CIVL Student: CHAN Wai Chung / ENVS Course: UROP1000, Summer Air quality closely correlates to health. A good indoor environment not only contributes to occupants' comfort but also safeguarding the health of the occupants. Recent SARS-CoV-2 cases have revealed that air is one of the transmission pathways of the disease. These cases show that the indoor air quality and ventilation is vital to reduce the concentration of airborne contaminants and control the spread of the disease. This paper aims to review different existing monitoring project and introduce how the SCAN project of HKUST could help to prevent the spread of disease. A safe indoor CO2 reference level would be estimated and the data obtained from the SCAN project could be used for ventilation assessment. Interdisciplinary Programs Office Division of Environment & Sustainability 167 Affordable, Sustainable Fashion!? Supervisor: SAUERWEIN Meike / ENVR Co-supervisor: MATUS Kira / ENVR Student: ALBUQUERQUE Laila Joy / EVMT Course: UROP1100, Summer Small luxury brands have begun offering sustainably produced clothing at high premium prices. Besides paying fair wages, better raw materials and waste treatment processes, the use of less harmful chemicals can be cost intensive and not a viable option for fast fashion brands or most consumer price points. Within this project, the technology available to produce sustainable fashion affordably will be explored by considering alternative innovations and practises, most notably those within the Hong Kong Research Institute Of Textiles And Apparel and the Handbook of Sustainable Luxury Textiles Volumes 1 & 2. By compiling the hotspots identified along the life cycle of fashion garments, barriers in greenifying the textile industry, maintaining production profitability, and scalability can be more clearly defined. Affordable, Sustainable Fashion!? Supervisor: SAUERWEIN Meike / ENVR Co-supervisor: MATUS Kira / ENVR Student: DEODHAR Aditi Parag / MGMT Course: UROP1000, Summer This research intends to explore the different existing innovations, technologies, and public policies that lead to the making and consumption of affordable and sustainable fashion in Hong Kong. It explores various lifecycle analyses of fabric production and identifies the key hotspots in the lifecycle that have the most impacts. It examines the feasibility of key innovations and technologies concerning production processes that reduce costs and also make the produced fabrics environmentally and socially responsible. Set targets and achieved outcomes of different public policies adopted by the Hong Kong Special Administrative Region (HKSAR) Government are examined. Finally, conclusions regarding the overall trends and observations are made, and future steps on how this research could be continued are discussed. Apps and Tools to Encourage Responsible, Sustainable Consumption Supervisor: SAUERWEIN Meike / ENVR Student: YU Wei Lin Whitney / EVMT Course: UROP1100, Summer Sustainability is a central facet to daily life, but education on such values very often fall short, especially in local high schools of Hong Kong. This UROP project aims to evaluate and explore methods to bring changes to sustainability education where it is needed. In order to accomplish this, research was conducted on how sustainability is currently taught, how students feel about it, and via interviews with individuals with relevant experience. Changes to how sustainability is taught can potentially be achieved in two directions: through bottom-up change by directly reaching out to build a teaching community, or top-down by providing input to the Curriculum Development Council groups in charge of related subjects. Interdisciplinary Programs Office Division of Environment & Sustainability 168 Deceptive Environmental Claims and Greenwashing Supervisor: SAUERWEIN Meike / ENVR Student: FU Jiayi / EVMT SUTEDJA Amrita Saraswati / ACCT Course: UROP1000, Summer UROP1000, Summer In today’s world, sustainability is no longer only an issue advocated by specific members of the community. As public’s awareness of sustainability increases, both producers and consumers are becoming more concerned about the presented behaviors of products and companies. Not all companies have successfully promoted their image through CSR, as they exaggerate the effect instead. The behavior is called ‘Greenwashing’, or ‘say more than do’. Social expectations and potential profits make the companies blindly try to cater to the trend, leading to the wrong path – greenwashing. Whatever the greenwashing motivation, in the long run, greenwashing brings more losses and risks than benefits. Producers should avoid falling into the trap. Our study focuses on the different criteria of definition on greenwashing; the negative impact of greenwashing on both sides – consumers and producers, as well as the political and economic environment; a few unnoticeable positive consequences. Finally, in the appendix, we made a Greenwashing Categorization table by Siano et al., TerraChoice, and Southern Weekly. Based on the research result, we made a survey determine the awareness of green and greenwashing among people, especially who has the power for daily necessities purchase. It is still on progress. Deceptive Environmental Claims and Greenwashing Supervisor: SAUERWEIN Meike / ENVR Student: ROMERO Sonia Mizuki / WBB Course: UROP1100, Fall Greenwashing, when companies deceptively overclaim the sustainability of their products or business practices, has become a rampant issue around the world over the past years, with Hong Kong being no exception. This article discusses the various ways to tackle Greenwashing and which strategy might be most effective in Hong Kong. Thirdparty regulation and consumer education was found to be time-consuming and ineffective. However, company internal regulation shows a lot of potential to reduce Greenwashing in Hong Kong and has already been successfully implemented abroad by companies like Walmart. Although tackling the problem of Greenwashing is a long, arduous and multi-faceted task, there is definitely potential for improvement and success for Hong Kong in the future. Interdisciplinary Programs Office Division of Environment & Sustainability 169 Deceptive Environmental Claims and Greenwashing Supervisor: SAUERWEIN Meike / ENVR Student: XIAO Hongyu / RMBI Course: UROP2100, Fall As we all know, Hong Kong consumers are paying more and more attention to living a green life. Many people realize that buying green products can contribute to the environment and make their own lives healthier. However, some companies and big stores are using these thoughts that consumers are willing to pay more money for green products. Those manufacturers produce a lot of goods with misleading advertisement, and some of strategies are called green washing. For example, some Fast-moving consumer goods (FMCG) companies put a logo which is very similar to an ISO certified logo on their packages, but they haven't really obtained the qualification. Due to the immature market and not sound law system, many companies get profit from greenwashing. the authorities also do not have clear rules nor guidance on how to punish or discourage false claims on product labels, many companies get profits with green washing. In this semester, we developed a questionnaire to research consumer behavior, and explore people’s knowledge on sustainable products, awareness of greenwashing and how their perception and actions may be affected depending on after they know about the existence of the green washing. This survey aims to ultimately give suggestions to different stakeholders what should be done to promote the further development of the sustainable product market in Hong Kong and avoid greenwashing. Deceptive Environmental Claims and Greenwashing Supervisor: SAUERWEIN Meike / ENVR Student: YEUNG Ho Fan Marco / EVMT Course: UROP1100, Fall An increasing number of consumers in Hong Kong show increasing awareness on sustainability concepts in recent years. To achieve sustainable lifestyle, green-claimed consumer goods are being preferred by such consumers and related markets have emerged rapidly. However, not all products that are labelled to be more sustainable than their conventional competitors really are. Deceptive environmental claims and labels on products confuse customers from truly sustainable products, and local consumers may find it difficult or lack of willingness to identify the credibility of green products they buy. Due to insufficient verification from both the corporate side and the government, the current consumption pattern induces “greenwashing” products be on sale in the consumer goods market in Hong Kong. This research investigates local consumers’ behavior towards sustainable and greenwashed products. Through the developed questionnaire we aim to investigate consumers’ ability and willingness in identifying greenwashing products. A hypothesis of 4 different consumer types ranging from (1) unwilling to buy, (2) consumers willing to buy sustainable goods but unable to identify truthfulness of claims (3) consumers willing to buy sustainable goods but unwilling to make and effort to identify truthfulness of claims, to (4) consumers willing to buy green and able to identify and willing to avoid green washing. Interdisciplinary Programs Office Division of Environment & Sustainability 170 Life Cycle Approaches for Sustainable Buildings and Infrastructures Supervisor: SAUERWEIN Meike / ENVR Co-supervisor: BOOKHART Davis Boyd / ENVR BROWN, David James / ENVR Student: CHAN Chun Yu / SENG Course: UROP1000, Summer The first part of this report will focus on the building certification. This report will first introduce China 3-star and then compare it with BEAM. The comparison can reflect the performance of Hong Kong building certification in LCA, benchmark, building services, green building material certification, and software recommendation. The second part of the report will focus on LCA tools. This report will introduce Athena Building Impact Estimator and then compare it with Hong Kong LCA tools, LCEA, and CIC CAT. This comparison can show the merits and drawbacks of Hong Kong LCA tools in transportation, life cycle information, and database. After that, this report will advise on helping Hong Kong LCA tools becoming more applicable to Hong Kong. Finally, this report will give suggestions for LCA promotion and education. Life Cycle Approaches for Sustainable Buildings and Infrastructures Supervisor: SAUERWEIN Meike / ENVR Co-supervisor: BOOKHART Davis Boyd / ENVR BROWN, David James / ENVR Student: CHAWLA Anhad Singh / SENG Course: UROP1000, Summer This research report explores the role and current situation of building life cycle assessment in Hong Kong. The goals, stakeholders, scope and boundaries of the report are highlighted. Embedded carbon in buildings and its contribution to global warming is discussed and a need for LCA is established. Building certification tool used in Hong Kong - BEAM is analysed and its instructions for conducting Life Cycle Assessments are noted. The LCA tool designed specifically for Hong Kong - Carbon Assessment Tool (CAT) is looked at in-depth to analyse its utility and applicability in Hong Kong. It is compared to other LCA tools available online to assess key aspects, strengths and weaknesses. CAT is a work in progress and not yet ready to meet BEAM requirements but otherwise well designed for use in Hong Kong. A holistic review of all building rating systems and LCA software provides the basis of several future research questions. Several suggestions for building construction industry stakeholders are also provided. Life Cycle Approaches for Sustainable Buildings and Infrastructures Supervisor: SAUERWEIN Meike / ENVR Co-supervisor: BOOKHART Davis Boyd / ENVR BROWN, David James / ENVR Student: GOSWAMI Arnab / CEEV Course: UROP1100, Summer The project studies the LEED Rating System (American green building standard), and its relevance to life-cycle analysis. The various impact parameters for designing and constructing a building that are listed by LEED are considered, and a comparison with BEAM (Hong Kong building standard) is drawn. The second part of the paper examines the life-cycle analysis tool (LCEA), created by the EMSD (Hong Kong SAR Government), and its input and impact parameters. Further, a comparison of the tool is drawn out with other similar tools and aspects of improvement are discussed. Abstracts of UROP Projects 2019-20 INSTITUTES & CENTERS Institutes & Centers Institute for Emerging Market Studies 172 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 / SOSC Student: BERRADA Ghali / WBB Course: UROP1000, Summer My participation in this research initiative only constitutes the early stage of a larger scale project to understand the effects of the Belt & Road Initiative on trade, but it has already been an extremely rewarding experience. With the help of Dr. Tritto, I have been able to understand the basics of doing research, and I have also learned to appreciate the less-funny task of data gathering that comes before any information can be analyzed, but I must say this is a crucial step in any research initiative, as the accuracy of the information will determine the accuracy of the final results of any research project. I have not only learned a lot about the method, but I have also greatly improved my understanding of the content of that research. Those couple of months have allowed me to know more about the Belt & Road Initiative, and the extent of Chinese investments in the ASEAN region. It has also made me more sensitive to the potential risks involved in investing in emerging and frontier markets, especially risks related to possible countries defaulting, as seen with covid-19. More generally, I will do my best to grow my interest in emerging markets and stay curious about research initiatives. Trade and Investment under China's Belt and Road Initiative - Implications for Hong Kong Supervisor: TRITTO Angela / IEMS Co-supervisor: PARK Albert Francis / SOSC Student: HOURI Olivia / WBB Course: UROP1000, Summer The Belt and Road Initiative (BRI) was first introduced in 2013 as a means of connecting China with the rest of Asia and Europe. This would be achieved through increased economic cooperation and new investments to build infrastructure to facilitate trade. In this project, investments related to the initiative are categorized according to their sustainability. Variables such as country of investment and ownership type of the investing company are compared to the number and capital involved in sustainable and non-sustainable investments to measure their correlation. Trade and Investment under China's Belt and Road Initiative - Implications for Hong Kong Supervisor: TRITTO Angela / IEMS Co-supervisor: PARK Albert Francis / SOSC Student: TONG Yvonne Chung Sze / GCS Course: UROP1100, Spring The ‘One Belt, One Road’ Initiative, comprising the Silk Road Economic Belt and the 21st Century Maritime Silk Road, can play a pivotal role in the economic future of China and the ASEAN countries. The Initiative bears much potential to enhance collaboration on regional infrastructure, trade settlements, and investment. Yet, while such enhancement is critical to China and the ASEAN countries, there is a lack of database of the complete infrastructural project list in academia for reviewing the progress of such Initiative in aiding ASEAN economy and, subsequently, helping the thrive of regional infrastructure to facilitate the success of the Initiative. With regards to such purpose, this report will present the background of the Initiative and the situation in ASEAN countries, in order to provide a detailed walkthrough of the procedures required in through the establishment of a database of BRI-related news. Institutes & Centers Institute for Emerging Market Studies 173 Trade and Investment under China's Belt and Road Initiative - Implications for Hong Kong Supervisor: TRITTO Angela / IEMS Co-supervisor: PARK Albert Francis / SOSC Student: YUAN Wanyi / QSA Course: UROP1000, Summer The report examines the sustainability of trades and investments under China’s Belt and Road Initiative throughout the period of 2008-2018 among Southeast Asian countries. This report explores BRI investments overseas and their implications on sustainability issues. The main focus of the report is to analyze different types and ownership of investments in 10 ASEAN countries based on quantitative analysis, compare investment types during pre-BRI and post- BRI periods, and identify BRI’s largest challenges and opportunities for sustainable development. With China’s strong commitment to green development of its multitrillion-dollar trade initiatives, this report delivers important insights to the future of China’s BRI investments in global environmental governance. MTPC-G20385