KUN JIN

University of Michigan, Ann Arbor 734-***-**** adkiiv@r.postjobfree.com https://www.linkedin.com/in/kun-jin-1426b1126/ RESEARCH INTEREST

Game Theory, MechanismDesign,Optimization, Network Science, DataMining,Machine Learning, Data-driven Security, Graph Learning, Knowledge Graph, Deep Learning

EDUCATION

• Ph.D., Electrical and Computer Engineering, University of Michigan, Ann Arbor, (4.0/4.0) Sep 2017 to present

• M.S., Electrical and Computer Engineering, University of Michigan, Ann Arbor, (4.0/4.0) Sep 2017 to May 2020

• B.S., Information Science, and B.S., Economics, Peking University, Beijing, China Sep 2013 to Jul 2017 SKILLS

• Programming: Python, MATLAB, SQL, Linux command, PyTorch, C++, Spark and TensorFlow

• Software: LaTeX, Microsoft Word, Excel, PowerPoint, XShell, PuTTY, FileZilla, Anaconda, Adobe Acrobat and Prezi INTERN EXPERIENCE

Tencent, Shenzhen, China

Machine Learning Research Intern Oct 2018 to May 2019 1. Portrait Creation for Companies:

• Designed and generated a variety of labels for companies, including fields, sizes, structures and trends. This project involved employee network detection, network community detection, node embeddings and centrality measure- ment, detection of companies’ change of focuses using data from social network APPs.

• Responsible for database operations and queries using SQL and PySpark; data processing with Python and Spark; implementation of centrality computation, label propagation algorithm, clustering(K-means, DBSCAN) and graph- ical machine learning algorithms(deep learning, e.g., Node2Vec) for node embedding, with Python; performance analysis and collaborations with other colleagues(Python packages: scikit-learn, numpy, scipy, pandas, networkx, matplotlib, wordcloud)

• This project studied and labeled more than 50 China A-shares companies. Among them we had a significant dis- covery on Bilibili’s change of focus from live stream to games before their annual report announced it, which was valuable for investors. This analysis module was the first one based on social network data and was later used by AI+investment teams.

2. Algorithm Design for Knowledge Graph Link Predictions:

• Designed and implemented two algorithms(deep learning) based on TransE/R/H/D, ConvE and their variants

• Responsible for data preprocessing(C++), algorithm design and implementation(Python, PyTorch, Keras), hyperpa- rameter tuning, performance analysis, and documentation writing for other colleagues to use the codes

• This project led to two functional algorithms with better performances(0.05 to 0.1 increase in MRR, Hit@10,3,1 on WordNet-18RR/0.1 to 0.15 increase on FreeBase15k-237) than TransE/R/H/D. RESEARCH EXPERIENCE

University of Michigan, Ann Arbor

Ph.D., Graduate Student Research Assistant

1. Game Theory on Networks with Group Structure: (Multidisciplinary University Research Initiative, MURI)

• Studied the properties of Nash equilibrium(NE); designed computationally efficient(up to 99.9% CPU time saving) algorithms to find the NE by utilizing the structures; studied the group formations of agents. The studied models work for a myriad of real world applications including public good provision, policy design, social networks, pricing and security problems

• Responsible for theorem finding, theoretical reasoning, algorithm design and implementation and results analysis

• Collaboration with professors in other universities and research fields in economics, social science and computer science

• This project led to 1 publication on IEEE American Control Conference(ACC) 2020, 1 publication on AAAI 2021, 1 submission to IEEE Control and Decision Conference(CDC) 2021 and 1 journal paper submission to IEEE Transactions on Automatic Control.

2. Data Driven Security:

• Designed a machine learning based scanner using XGBoost(Python) that significantly improves the scanning effi- ciency on IPv4 addresses. Saves more than 80% of resources while receiving 99% of the responses. The machine learning module can work for a more general class of multi-label classification problems with label correlations

• Responsible for data processing, algorithm design and implementation, metric design and performance analysis

• This project led to 1 submission “Towards Bandwidth-Conserving Internet Scans: Intelligent and Adaptive Probing Using Machine Learning” under review to ACM SIGMETRICS 2021 and 1 patent Peking University, Beijing, China

Undergraduate Student Research Assistant

• Wireless channel modeling and simulation for unmanned aerial vehicles(UAV) using MATLAB

• This project led to 1 publication on IEEE VTC, Sidney, spring 2017 PUBLICATIONS (ALL AS FIRST AUTHOR)

• “Multi-Scale Games: Representing and Solving Games on Networks with Group Structure” by Kun Jin, Yevgeniy Vorob- eychik and Mingyan Liu on Association for the Advancement of Artificial Intelligence (AAAI) 2021(accepted)

• “Games on Networks with Community Structure: Existence, Uniqueness and Stability of Equilibria” by Kun Jin, Mo- hammad Mahdi Khalili and Mingyan Liu on American Control Conference (ACC) 2020, IEEE

• “ThreeDimensionalModeling and Space-Time Correlation for UAV Channel” by Kun Jin, Xiang Cheng, Xiaohu Ge and Xuefeng Yin on IEEE Vehicular Technology Conference (VTC), Sidney, spring 2017 AWARDS

• University of Michigan, College of Engineering Graduate Scholarship

• PKU EECS Department Undergrad Scholarships 2014, 2016

Contact this candidate