管理科学系学术讲座（5月22日）

时  间：2026年5月22日（周五） 10:30-11:30

地  点：管理学院思源楼524室

主 题：Black-Box Quantile Optimization Using Finite Difference-Based Gradient Approximation

主讲人：胡家翘 纽约州立大学石溪分校教授

主持人：胡建强 复旦大学管理学院教授

Abstract:We propose two new iterative multi-time scale stochastic gradient decent algorithms for solving quantile optimization problems under a general black-box setting. The first algorithm uses an appropriately modified finite-difference-based gradient estimator that requires 2d +1 samples of the black-box function per iteration of the algorithm, where d is the number of decision variables. The second algorithm employs a simultaneous-perturbation-based gradient estimator that uses only three samples for each iteration regardless of problem dimension. We show the almost sure convergence of both algorithms and establish their rates of convergence. Numerical results are also reported to illustrate and compare the performance of the algorithms with alternative methods.

Bio:Jiaqiao Hu is a Professor in the Department of Applied Mathematics and Statistics at the State University of New York, Stony Brook. He received the B.E. degree in automation from Shanghai Jiao Tong University, the M.S. degree in applied mathematics from the University of Maryland, Baltimore County, and the Ph.D. degree in electrical engineering from the University of Maryland, College Park. His research interests include Markov Decision Processes, simulation optimization, and stochastic modeling and analysis. He has been on the editorial boards of Operations Research, IISE Transactions, and Journal of Systems & Management.