BackdoorBench: A Comprehensive Benchmark of Backdoor Learning
Wu, Baoyuan, et al. "BackdoorBench: A Comprehensive Benchmark of Backdoor Learning." NeurIPS 2022 Track Datasets and Benchmarks.
CV
Education
- B.S. in Electronic Information Engineering, The Chinese University of Hong Kong, Shenzhen, 2020
- Ph.D. Candidate in Data Science, The Chinese University of Hong Kong, Shenzhen, 2020 - Present
Skills
- Programming Languages: Python, MATLAB, C/C++, Julia, R
- Development Environments: PyTorch, TensorFlow, Pandas, Linux
- Languages: English, Fluent. Chinese, Mother Tongue
Publications
Mean Parity Fair Regression in RKHS
Wei, Shaokui, et al. "Mean Parity Fair Regression in RKHS" AISTATS 2023.
Enhancing Fine-Tuning Based Backdoor Defense with Sharpness-Aware Minimization
Zhu, Mingli, et al. "Enhancing Fine-Tuning Based Backdoor Defense with Sharpness-Aware Minimization." 2023 International Conference on Computer Vision"
Neural Polarizer: A Lightweight and Effective Backdoor Defense via Purifying Poisoned Features
Zhu, Mingli, Shaokui Wei (co-first author), et al. "Neural Polarizer: A Lightweight and Effective Backdoor Defense via Purifying Poisoned Features." Thirty-seventh Conference on Neural Information Processing Systems. NeurIPS 2023.
Shared Adversarial Unlearning: Backdoor Mitigation by Unlearning Shared Adversarial Examples
Wei, Shaokui, et al. "Shared Adversarial Unlearning: Backdoor Mitigation by Unlearning Shared Adversarial Examples." Thirty-seventh Conference on Neural Information Processing Systems. NeurIPS 2023.
VDC: Versatile Data Cleanser for Detecting Dirty Samples via Visual-Linguistic Inconsistency
Zhu, Zihao, et al. "VDC: Versatile Data Cleanser for Detecting Dirty Samples via Visual-Linguistic Inconsistency." ICLR 2024.
Mitigating Backdoor Attack by Injecting Proactive Defensive Backdoor
Wei, Shaokui, et al. "Mitigating Backdoor Attack by Injecting Proactive Defensive Backdoor." Thirty-eighth Conference on Neural Information Processing Systems. NeurIPS 2024.
Unveiling and Mitigating Backdoor Vulnerabilities based on Unlearning Weight Changes and Backdoor Activeness
Lin, Weilin, et al. "Unveiling and Mitigating Backdoor Vulnerabilities based on Unlearning Weight Changes and Backdoor Activeness." Thirty-eighth Conference on Neural Information Processing Systems. NeurIPS 2024.
Talks
Logarithmic Regret Algorithms for Online Convex Optimization
Weekly Talk at CUHKSZ, Shenzhen, China
Solving Mixed Integer Programs Using Neural Networks
Weekly Talk at CUHKSZ, Shenzhen, China
Bias and Fairness in Machine Learning
Weekly Talk at CUHKSZ, Shenzhen, China
Generative Models
Weekly Talk at CUHKSZ, Shenzhen, China
Energy-based Learning for Cooperative Games in Valuation Problems
Weekly Talk at CUHKSZ, Shenzhen, China
On the Tradeoff between Fairness and Robustness
Weekly Talk at CUHKSZ, Shenzhen, China
Distributionally Robust Learning
Weekly Talk at CUHKSZ, Shenzhen, China
Loss Surfaces, Mode Connectivity and Model Robustness
Weekly Talk at CUHKSZ, Shenzhen, China
Improving Robustness using Generated Data
Weekly Talk at CUHKSZ, Shenzhen, China
Sharpness-Aware Minimization
Weekly Talk at CUHKSZ, Shenzhen, China
Structured Pruning of Deep Convolutional Neural Networks
Weekly Talk at CUHKSZ, Shenzhen, China
Tutorial on Backdoor Learning: Recent Advances and Future Trends
Tutorial at ICCV 2023, Pairs, France
Advanced Works in ICCV 2023: Backdoor Learning I
Weekly Talk at CUHKSZ, Shenzhen, China
Backdoor Mitigation via Adversarial Training Techniques
Talk at AIRS, Shenzhen, China