Kai (James) Zhang | GIBDL, SRIBD

About Me

Kai Zhang is a research data engineer at Shenzhen Research Institute of Big Data (SRIBD). He is affiliated to the Government and Industrial Big Data Lab where he primarily works on delivering insight and impact for government and enterprises through a wide range of flexible support models, providing ad hoc, deeply transformational, and ongoing analytics architecture and solutions. Before joining SRIBD, Kai received his Master’s degree in Biostatistics from Cornell University, and a Bachelor’s degree in Statistics from Chinese University of Hong Kong, Shenzhen.

Research Interests

Statistical Learning: time series, quantitative finance, interpretable machine learning
Causal Inference: econometrics, targeted learning, experimental design and analysis

Projects

Comprehensive Python Framework for Factor Investment (Supervised by Prof. Tao Shu at CUHKSZ)
This project is aimed at providing a comprehensive software framework for factor investment, including data interface, factors library, factor analysis, and strategy backtesting. Data interface module supports storing and importing of financial data. Factor analyzer module provides performance analysis of predictive stock factor, including Fama-MacBeth regression, time series regression, barra regression, informaction coefficient analysis and fractile analysis, etc. Backtester module provides vectorized backtesting framework for investment strategy.
Automated Project Auditing System Based on Semantic Text Matching Algorithms (Supervised by Prof. Tao Shu at CUHKSZ)
This project developed text semantic matching metrics based on WordNet lexical database with Wu Palmer similarity and Word2Vec algorithm with cosine similarity, providing Shenzhen Science and Technology Innovation Commision with a solution to audit the research outputs of the state-funded projects in batches.
Cell Type Identification of Mouse Retinal Pigment Epithelium from ScRNA-seq Data (Supervised by Prof. Xi Kathy Zhou at Cornell University)
This project used unsupervised learning methods to identify the cellular composition and activities of mouse retinal pigment epithelium (RPE) from single-cell transcriptomics, and further to detect cell-to-cell variation by statistical methods. We firstly processed raw sequence data into a high-quality expression data following Scater workflow, including rigorous pre-processing, quality control, feature selection and PCA dimensionality reduction. Then, multiple clustering algorithms were performed and compared on cells to identify RPE cell types, including K-means, DBSCAN and SNN graph-based clustering. Finally, potential gene markers were detected for each cell class using pairwise t-test with BH adjustment for multiple comparisons.
Statistical Mediation Analysis on Components of Frailty with an Ordinal Outcome (Supervised by Prof. Arindam RoyChoudhury at Cornell University)
This project investigated into Racho Bernado study, an observational cohort study, regarding mechanism of frailty in older adults. Evidence of mediation relationship among components of frailty (strength, body composition, and body performance) is provided by traditional Baron and Kenny’s three-steps method based on proportional odds model.

News

[Feb. 2020] Our paper about incremental learning is accepted to CVPR 2020.
[Feb. 2020] We will host the ACM Multimedia Asia 2020 conference in Singapore!
[Sept. 2019] Our paper about few-shot learning is accepted to NeurIPS 2019.
[Mar. 2019] Our paper about few-shot learning is accepted to CVPR 2019.

Publications

Mnemonics Training: Multi-Class Incremental Learning without Forgetting
Yaoyao Liu, Yuting Su, An-An Liu, Bernt Schiele, Qianru Sun
IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2020.
[PDF] [Code] Oral Presentation
Learning to Self-Train for Semi-Supervised Few-Shot Classification
Xinzhe Li, Qianru Sun, Yaoyao Liu, Shibao Zheng, Qin Zhou, Tat-Seng Chua, Bernt Schiele
33rd Conference on Neural Information Processing Systems. NeurIPS 2019.
[PDF] [Code]
Meta-Transfer Learning for Few-Shot Learning
Qianru Sun*, Yaoyao Liu*, Tat-Seng Chua, Bernt Schiele
IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2019.
[PDF] [Code] [Project]

Services

Conference Reviewers: NeurIPS 2020, CVPR 2020.
Journal Reviewers: T-PAMI, IJCV.