北京科学与工程计算研究院学术报告之六十三

报告人/Speaker: 宋恩彬(四川大学)


报告题目/Title:  Block Coordinate Descent Almost Surely Converges to a Stationary Point Satisfying the Second-order Necessary Condition


时间/Date & Time: June 19, 2018, 9:30-10:30


地点/Location: 理科楼M842报告厅


报告摘要/Abstract:

Given a non-convex twice continuously differentiable cost function with Lipschitz continuous gradient, we prove that all of the block coordinate gradient descent, block mirror descent and proximal block coordinate descent methods converge to stationary points satisfying the second order necessary condition, almost surely with random initialization. All our results are ascribed to the center-stable manifold theorem and Ostrowski’s lemma.


报告人简介/About the speaker:

Enbin Song received the degree of the bachelor of science from Shandong Normal University in 2002. In 2007, he received the Ph.D. degree from Sichuan University, China. He was a visiting scholar at School of Electronic and Information Engineering, Xi’an Jiao Tong University from September 2007 to January 2008. From February 2008 to Nov. 2009, he was a postdoctoral at Department of Computer Science, Sichuan University and from April 2010 to May 2011, he was a postdoctoral at Department of Electrical and Computer Engineering, University of Minnesota. From November 2012 to December 2012, he is a visit scholar with Institute of Computational Mathematics and Scientific/Engineering Computing of the Chinese Academy of Sciences and from December 2013 to February 2014, he was a visit scholar with the Department of Systems Engineering and Engineering Management, The Chinese University of Hong Kong. He has been with Department of Mathematics, Sichuan University since November 2009 as Associate Professor (November 2009--June 2014) and has been a Professor since July 2014.
His main research interests are in information processing, nonlinear optimization and large-scale optimization methods for big data problem (both algorithmic design and theory) and its application, including signal processing and statistics, machine learning, the theory of multi-user wireless communications, particularly, with an emphasis on the application of optimization techniques to multi-sensor estimation and decision fusion theory.