Southwest Jiaotong University School of Mathematics


首页  >  学术科研  >  学术交流  >  正文

Efficient DC Algorithm for Nonconvex Nonsmooth Optimization Problems

来源:   作者:     日期:2019-08-30 15:47:06   点击数:  

报告题目:Efficient DC Algorithm for Nonconvex Nonsmooth Optimization Problems

报告人:Akiko Takeda, 武田朗子 教授 (University of Tokyo/RIKEN, Japan






 There are many applications in signal processing, machine learning and operations research that seek sparse solutions by adopting the L0-norm constraint. We address the minimization of a smooth objective function under an L0-norm constraint and simple constraints. Some efficient algorithms are available when the problem has no constraints except the L0-norm constraint, but they often become inefficient when the problem has additional constraints. We reformulate the problem by employing a new DC (difference of two convex functions) representation of the L0-norm constraint, so that DC algorithm can retain the efficiency by boiling down its subproblems to the projection operation onto a simple set. We will also discuss the extension of the approach to more general problems whose objective is the sum of a smooth function and proper closed possibly nonsmooth functions (whose proximal mappings are easy to compute).

 This talk is based on joint works with K. Tono, J. Goto, T. Liu and T.K. Pong.


 Akiko Takeda received the Bachelor of Engineering degree and the Master of Engineering degree from Keio University, Japan, in 1996 and 1998, respectively, and the Doctor of Science degree in information science from the Tokyo Institute of Technology, Japan, in 2001. She was with Corporate R&D Center at Toshiba Corporation from 2001 to 2003, with Tokyo Institute of Technology from 2003 to 2008, with Keio University from 2008 to 2013, with University of Tokyo from 2013 to 2016 and with the Institute of Statistical Mathematics from 2016 to 2018. She is currently a Professor with the Department of Creative Informatics, University of Tokyo (2018-present), and the Team Leader of Continuous Optimization Team at RIKEN Center for Advanced Intelligence Project, Tokyo, Japan (2016-present). Her current research interests include solution methods for decision making problems under uncertainty and nonconvex optimization problems, which appear in financial engineering, machine learning, and energy systems.

 She is a recipient of the 7th Information Technology Young Investigator Award (2008). She also received the 6th Research Award (2016) of the Operations Research Society of Japan (ORSJ). She is a Fellow of ORSJ since 2017. She serves as a Local Organizing Committee Member of the 5th International Conference on Continuous Optimization (ICCOPT 2016 Tokyo), and Publication Chair for the 22nd International Conference on Artificial Intelligence and Statistics (AISTATS 2019).