Decision Making in Stochastic Dynamic Systems: Principled Data-driven Methods

Date: 2024/08/02 - 2024/08/02

Academic Seminar: Decision Making in Stochastic Dynamic Systems: Principled Data-driven Methods

Speaker: Dengwang Tang, postdoctoral researcher, University of Southern California (USC)

Time: 13:30-14:30, August 2, 2024 (Beijing Time)

Location: Room 454, JI Long Bin Building

Abstract

Stochastic dynamic system models have widespread applications in settings such as robotics, autonomous vehicles, power grids, and wireless communications. Today’s dynamic systems are often large, very uncertain, decentralized, with many unknown parameters. Decision making in such systems presents significant challenges, since many traditional methods, developed with a focus on much smaller systems, either cannot be applied or can be computationally intensive or data inefficient. My research aims at developing new theories and algorithms to meet the challenges of data-driven decision making in today’s large scale multi-agent systems. In this talk, I will discuss my contributions in (1) reinforcement learning algorithms and multi-armed bandits, where I devise novel, principled data-driven algorithms; (2) information states for multi-agent dynamic games, where I develop structural results and provide new insights; (3) planning for multi-agent dynamic systems, where I develop efficient algorithms for multi-agent teams; (4) dispatching algorithms for multi-server cloud computing centers, where I proposed and systematically analyzed a low-randomness dispatching algorithm. Finally, I will outline my expertise and future research plans that will serve my ultimate goal of enabling well-functioning large-scale dynamic systems.

Biography

Dengwang Tang is a postdoctoral researcher at University of Southern California (USC). He earned his Ph.D. in Electrical and Computer Engineering (2021), M.S. in Mathematics (2021), and M.S. in Electrical and Computer Engineering (2018) all from University of Michigan (UM), Ann Arbor. He obtained B.S.E degrees in Electrical and Computer Engineering at Shanghai Jiao Tong University (SJTU) and in Computer Engineering from University of Michigan, Ann Arbor in 2016 through the dual degree program at the UM-SJTU Joint Institute. Prior to joining USC he was a postdoctoral researcher at University of California, Berkeley. His research interests involve statistical learning algorithms in stochastic dynamic systems, multi-armed bandits, multi-agent systems, and game theory, with applications to robotics, autonomous vehicles, recommender systems, and beyond.