| Time: | February 7, 2017 |
|---|---|
| Download as iCal: |
|
Prof. Pontus Giselsson
Dept. of Automatic Control
Lund University, Sweden
Tuesday, 2017-02-07 16:00
IST-Seminar-Room V9.22 - Pfaffenwaldring 9 - Campus Stuttgart-Vaihingen
Abstract
Many popular first order algorithms for convex optimization, such as forward-backward splitting, Douglas-Rachford splitting, and the alternating direction method of multipliers (ADMM), can be formulated as averaged iteration of a nonexpansive mapping. In this paper we propose a line search for averaged iteration that preserves the theoretical convergence guarantee, while often accelerating practical convergence. We discuss several general cases in which the additional computational cost of the line search is modest compared to the savings obtained.
Biographical Information
Pontus Giselsson is an Associate Professor at the Department of Automatic Control at Lund University, Sweden. His current research interests include mathematical optimization and its applications in, e.g., control, signal processing, and machine learning. He received a M.Sc. degree from Lund University in 2006 and a Ph.D. degree from Lund University in 2012. During 2013 and 2014, he was a postdoc at Stanford University. In 2012, he received the Young Author Price at an IFAC Symposium, in 2013 he was a finalist for the Best Student Paper Award at the American Control Conference, in 2014 he received the Young Author Price at the IFAC World Congress, and in 2015 he received the Ingvar Carlsson Award from the Swedish Foundation for Strategic Research.