主题：Reinforcement Learning in Mean Field Games: the Pitfalls and Promises

报告人：Niao HE, Assistant Professor, Department of Computer Science, ETH Zurich

主持人：Ming YAN, Associate Professor & Assistant Dean (Undergraduate Affairs), School of Data Science, CUHK-Shenzhen

日期：23 January (Tuesday), 2024

时间：11:00 AM - 12:00 PM, Beijing Time

形式：Hybrid

地点：103 Meeting Room, Daoyuan Building

SDS视频号直播：

语言：English

摘要：

Amidst its diverse applications, multi-agent reinforcement learning (MARL) stands as a formidable challenge in general due to the “curse of many agents”. Mean-field games (MFGs) have merged as a “tractable” surrogate for modeling game dynamics involving a large population of interacting decision-makers. Despite the widespread use of reinforcement learning techniques in recent literature to address MFGs, several foundational questions persist:

(1) When are MFGs good approximations of MARL problems?

(2) When does learning MFGs exhibit computational or statistical tractability?

(3) Can Nash equilibrium be achieved through independent learning?

This talk will explore these crucial dimensions of learning MFGs, with a mix of positive and negative findings, shedding light on their fundamental limits.

简介：

Niao He is an Assistant Professor in the Department of Computer Science at ETH Zurich, where she leads the Optimization and Decision Intelligence (ODI) Group. She is also an ELLIS Scholar and a core faculty member of ETH AI Center, ETH-Max Planck Center of Learning Systems, and ETH Foundations of Data Science. Previously, she was an assistant professor at the University of Illinois at Urbana-Champaign from 2016 to 2020. Before that, she received her Ph.D. degree in Operations Research from Georgia Institute of Technology in 2015. Her research interests lie in large-scale optimization and reinforcement learning, with a primary focus on theoretical and algorithmic foundations for principled, scalable, and trustworthy decision intelligence.