时间:2025年3月28日(星期五)10:00-11:00
地点:E4-233
主持人: 理论科学研究院 周珍楠
主讲人: 澳门大学 徐礼虎
主讲人简介:徐礼虎博士现任澳门大学副教授,2014年加入该校至今。他于2008年获英国帝国理工学院博士学位,先后在德国波恩大学(2008)、荷兰埃因霍温理工大学(2008-2010)及德国柏林工业大学(2010-2011)从事博士后研究工作。2011至2013年间,他在伦敦布鲁内尔大学担任讲师(永久职位,等同助理教授)。其研究方向为应用概率及其在统计理论中的应用,具体包含以下研究课题:Stein方法、随机逼近理论、大偏差原理、Malliavin微积分及稳健统计估计,已在《应用概率年刊》《统计年刊》《伯努利》《泛函分析杂志》《概率论及相关领域》等国际期刊发表论文50余篇。
讲座主题:Recent advances in Stein's method and related stochastic approximations
讲座摘要: We review recent advances in Stein's method and the related stochastic approximation theory, emphasizing three interconnected themes: 1) Advancing stable law approximations via Stein's method, with explicit error bounds for systems exhibiting heavy-tailed behavior; 2) Establishing a new Stein's method for diffusion approximation, which enables a tractable steady-state analysis of queueing systems through approximations of stationary measures for stochastic differential equations. 3) Establishing a unified Markovian framework for stochastic approximations of stochastic optimization algorithms—including SGD, SVRG, and momentum-based variants—to enable comparative analysis of their convergence dynamics. By synthesizing probabilistic tools with applications in optimization and queueing theory, we demonstrate the unifying role of stochastic approximations in bridging theoretical innovation with practical challenges.