管理科学与工程高端论坛:刘光梧教授

时间:2020-09-04  阅读量:484次

报告题目:An Upper Confidence Bound Approach to Estimating the Maximum Mean

报告时间:2020923日(周三)下午2:00-4:00

报告平台:腾讯会议 (会议ID709 968 812)

主办单位:管理科学与工程学院大数据管理与优化研究中心

 

【报告人简介】

Prof. Guangwu Liu is currently the acting head of Department of Management Sciences, College of Business at City University of Hong Kong. His research interests include stochastic simulation, machine learning, business analytics, financial engineering and risk management. He has published in top journals of the field, including Operations Research, Management Science, INFORMS Journal on Computing, and ACM Transactions on Modeling and Computer Simulation. He has been serving as an associate editor of Naval Research Logistics since 2018.

【报告摘要】

Estimating the maximum mean of a number of stochastic systems finds a variety of applications in both management science and machine learning, ranging from financial risk measurement and Markov decision processes to reinforcement learning and Monte Carlo tree search. In this work, we study the estimation of the maximum mean under a generalized upper confidence bound (UCB) framework where the sampling budget is sequentially allocated to one of the systems. We study in depth the existing Grand Average (GA) estimator and propose a new Largest-Size Average (LSA) estimator. Specifically, we establish statistical guarantees, including strong consistency, central limit theorems (CLTs), and asymptotic mean squared errors for both estimators, which are new to the literature. We further construct asymptotically valid confidence intervals based on CLTs. Statistical efficiency of the resulting point and interval estimators is demonstrated via numerical examples.


电话:0411-84710475 邮编:116025 地址:大连市沙河口区尖山街217号
Copyright © 2014-2019 管理科学与工程学院

0.489535s