Academy of Mathematics and Systems Science, CAS
Colloquia & Seminars

Speaker:	陈亮，湖南大学
Inviter:
Title:	Decomposition and Approximation Approaches for Large-Scale Convex Composite Programming, Part II
Time & Venue:	2022.01.17 14：00-17：00 腾讯会议 ID: 355539240 密码：2022
Abstract:	In recent years, convex composite programming (CCP) has been extensively studied from many different angles. In this short course, we mainly introduce some decomposition and approximation approaches and techniques for handling the computational burdens from solving large-scale CCP problems. We will introduce a a unified algorithmic framework for CCP, which distills and unifes all the practical techniques that were constructively exploited in its various precursors for the purpose of improving the theoretical guarantee and the computational efficiency of multi?block alternating direction method of multipliers (ADMM). We also provide a proximal augmented Lagrangian interpretation of the ADMM, which may explain why it often converges slowly after the initial iterations (so that higher order algorithms are necessary).