Robust and Adaptive Algorithm Design in Online Learning

Mr. ZHANG Mengxiao
Ph.D. student
Computer Science Department
University of Southern California

Meeting ID: 986 7703 7612 // Passcode: 202400(Students must login with CUHK account, i.e.,, for valid attendance record)

Online learning, also known as sequential decision making, has been serving as an important component in modern machine learning with wide applications including large language models, auction, and clinic trials.  Different from the conventional offline learning framework where the data typically come from a certain distribution, the online learning environment can be more complicated.  The focus of my research is then to design robust and adaptive learning algorithms, which can perform well in both the benign and the possibly adversarial environments.  In this talk, I will discuss my works on how to apply modern techniques to handle various online learning problems with adaptivity and robustness guarantees. Furthermore, I will provide insights into the application of these algorithms in real-world contexts.  Finally, I will give an outline of my future work from both the theoretical and the empirical aspects.


Mengxiao Zhang is a Computer Science Ph.D. student at University of Southern California advised by Prof. Haipeng Luo. His research is about designing practical and adaptive machine learning algorithms with strong theoretical guarantees, with a focus on general sequential learning problems, including online learning, bandit problems, reinforcement learning, game theory and various operations and revenue management applications. He has published papers in top-tier machine learning conferences (ICML, NeurIPS, ICLR, AISTATS) and learning theory conferences (COLT, ALT), including oral presentations. Previously, he has interned with Microsoft Research and Amazon, and received a B.S. in School of EECS from Peking University.


Mr. WONG O-Bong (

Ms. FUNG Wing Chi Mary (


Apr 08, 2024


10:00 am - 11:00 am



Comments are closed.