Abstract

Abstract: Shannon’s information measures played a fundamental role in the study of information storage and communication. In recent years, however, such measures also emerged naturally in the study of machine learning theory and algorithms. In this presentation, we present new results on three such problems: 1) the study of multi-armed bandits when there exists heterogeneity in the reward variances, 2) policy optimization for the Markov decision process with multiple reward functions, and 3) the study of machine learning generalization error bounds. It is shown that information measures can also naturally capture effects in learning theory and algorithms that are not directly related to information storage and communications, and we expect them to play pivotal roles in the study of machine learning algorithms in the future.

Bio: Dr. Chao Tian received the B.E. degree in Electronic Engineering from Tsinghua University, Beijing, China, and the M.S. and Ph. D. degrees in Electrical and Computer Engineering from Cornell University, Ithaca, NY. Dr. Tian was a postdoctoral researcher at Ecole Polytechnique Federale de Lausanne (EPFL), then a member of technical staff—research at AT&T Labs—Research, an Associate Professor in the Department of Electrical Engineering and Computer Science at the University of Tennessee Knoxville, and is now an Associate Professor in the Department of Electrical and Computer Engineering at Texas A&M University. His authored and co-authored papers received the 2014 IEEE Data Storage Best Paper Award, the 2017 IEEE Jack Keil Wolf ISIT Student Paper Award, and the 2020-2021 IEEE Data Storage Best Student Paper Award. He was an Associate Editor for the IEEE Signal Processing Letters 2012-2014, an Editor for the IEEE Transactions on Communications 2016-2021, and an Associate Editor for the IEEE Transactions on Information Theory 2021-2023. He is a general co-chair of the 2024 IEEE Information Theory Workshop.