A study of the classical consensus-based optimization model
数学专题报告
报告题目(Title):A study of the classical consensus-based optimization model
报告人(Speaker):黄辉 (奥地利格拉茨大学)
地点(Place):后主楼1124
时间(Time):2025年7月30日 星期三 16:00-17:00
邀请人(Inviter):袁迪凡
报告摘要
Consensus-based optimization (CBO) is a relatively recent class of particle-based methods for global optimization, inspired by collective behavior in nature and rooted in stochastic dynamics. In this talk, we introduce the classical CBO model, where a swarm of agents iteratively updates their positions under the influence of attraction to a weighted consensus point and random perturbations. We present the mathematical formulation of the CBO model and discuss its key features. In particular, we shall establish its global convergence in mean-field law toward the global minimizer of a given objective function under suitable assumptions.
主讲人简介
黄辉,奥地利格拉茨大学研究助理,于2017年毕业于清华大学数学科学系,师从简怀玉教授和刘建国教授。博士毕业后曾先后在西蒙弗雷泽大学,慕尼黑工业大学以及卡尔加里大学从事博士后研究工作。主要研究方向为多粒子模型及其所对应的偏微分方程,以及他们在最优化领域中的应用。目前已发表文章30余篇,其中包括 J. Mach. Learn. Res., SIAM J. Optim. , SIAM J. Control Optim., Math. Comput., SIAM J. Math. Anal., M3AS 以及JDE等国际一流期刊。