报告题目：Linearized proximal algorithms with adaptive stepsizesfor convex composite optimization with applications

报 告 人：李冲教授，浙江大学

报告时间：2023年4月13日15:00-16:00

报告地点：21-410

报告摘要：In this talk, we continue to study the problem of numerically solving convexcomposite optimizations. Linearized proximal algorithms (LPA) with adaptive stepsizes forsolving the convex composite optimization problem are proposed. Local and/or global convergenceproperties of the proposed algorithms are explored, and their superlinear/quadraticconvergence results are established under the assumptions of local weak sharp minima andthe quasi-regularity condition. Our proposed algorithms, compared with the LPA with theconstant stepsize, have the advantages of suiting for wider range of problems and of employinghigher convergence rates. We apply the LPA with adaptive stepsizes to solve thewirelesssensor network localization problem, and the numerical results show that the LPAwithadaptive stepsizes can solve this problem more e_ciently and stable than the LPA withthe constant stepsize or other algorithms.

报告人简介：李冲，浙江大学数学系教授，博士生导师。主要从事非线性优化理论与计算、数值泛函分析、数值代数、稀疏优化及其应用、机器学习等领域的研究。先后主持国家自然科学基金及省部级项目等近二十项，出版专著1部，在SCI期刊上发表论文近200篇， 特别是在优化理论和计算数学的顶级刊物SIAM J Optim., Math. Program，SIAM J. Control Optim.以及SIAM J.Numer. Anal上发表论文近30篇。

邀请人：沈卫平

欢迎感兴趣的教师和研究生参加！