AQFC2015

Randomized block proximal gradient methods for a class of structured nonlinear programming

 


Department of Systems Engineering and Engineering Management
The Chinese University of Hong Kong

Title: Randomized block proximal gradient methods for a class of structured nonlinear programming
 
Speaker: Dr. Zhaosong Lu,
Simon Fraser University
 
Date: July 18, 2014 (Friday)
 
Time: 4:30 PM to 5:30 PM
 
Venue: Room 803
       William M.W. Mong Engineering Building (ERB)
       (Engineering Building Complex Phase 2)
       The Chinese University of Hong Kong.

Short Biography of the speaker:
Zhaosong Lu is an Associate Professor of Mathematics and an associate faculty member in Statistics and Actuarial Science at Simon Fraser University, Canada. He received his PhD in Operations Research from the school of Industrial and Systems Engineering of Georgia Institute of Technology in 2005 under the supervision of Dr. Renato D. C. Monteiro and Dr. Arkadi S. Nemirovski. He was a Zeev Nehari Visiting Assistant Professor of Mathematical Sciences at Carnegie Mellon University during 2005-2006. He also held Visiting Associate Professor positions at Texas A&M University and Arizona State University during 2012-2013. Dr. Lu's research interests include theory and algorithms for continuous optimization, and applications in data mining, finance, statistics, machine learning, image processing, engineering design, and decision-making under uncertainty. He was a finalist of 2005 INFORMS George Nicholson Best Student Paper Competition for the work on first-order methods for solving large-scale well-structured semidefinite programming. He has published numerous papers in major journals of his research areas such as Mathematical Programming, SIAM Journal on Optimization, INFORMS Journal on Computing, SIAM Journal on Matrix Analysis and Applications, Journal of the Royal Statistical Society, and ASME Journal of Mechanical Design.
 
Abstract:
Nowadays randomized block proximal gradient descent (RBPG) methods become a prevalent tool for solving large-scale optimization problems arising in machine learning, compressed sensing, image and signal processing. In the first part of this talk we study a randomized monotone block gradient method for minimizing the sum of a smooth convex function and a block-separable convex function. We present some new results on rate of convergence and high-probability type of iteration complexity for this method. We also propose an accelerated RBPG method and establish its rate of convergence. We also present some computational results. In the second part we propose a randomized nonmonotone block proximal gradient (RNBPG) method for minimizing the sum of a smooth (possibly nonconvex) function and a block-separable (possibly nonconvex nonsmooth) function. Under some assumptions, we establish its global convergence and rate of convergence. We also present some computational results demonstrating that our method substantially outperform the RBPG method proposed by Richtarik and Takac (2012).
 
                      Everyone is invited to attend the talk.
 
The talk will be hosted by:
       Prof. Shiqian Ma,
       Department of Systems Engineering and Engineering Management,
       The Chinese University of Hong Kong,
       E-mail: sqma@se.cuhk.edu.hk
       Telephone Number: (852) 3943-8240
 
For general enquiries, please contact the student coordinator:
       Andy Chung,
       Department of Systems Engineering and Engineering Management,
       The Chinese University of Hong Kong,
       E-mail: oychung@se.cuhk.edu.hk
 
SEEM-5202 Website: http://seminar.se.cuhk.edu.hk
Date: 
Friday, July 18, 2014 - 08:30 to 09:30