POSTPONED - CORMSIS Seminar - Chee Khian Sim Event

Abstract: Active research on solving optimisation problems using first order methods is currently being carried out by many researchers in the optimisation community.  This research can be applied to areas such as machine learning, hence it is also carried out by researchers working on machine learning.   Fast iterative shrinkage-thresholding algorithm (FISTA), a first order algorithm, was introduced by A. Beck and M. Teboulle in 2009 to solve convex composite optimisation problems.  It has been extended more recently to solve nonconvex composite optimisation (NCO) problems as well.    In this talk, I will describe two variants of FISTA to solve NCO problems.  I will provide iteration complexity results for these FISTA variants to find an approximate solution to an NCO problem.   I will also give numerical results using these variants to solve instances of NCO problem, comparing their performance with that of other known first order algorithms.

Brief biography: Dr. Sim is a Senior Lecturer with the School of Mathematics and Physics, University of Portsmouth.  He obtained his Ph.D., specialising in optimisation, from the Department of Mathematics, National University of Singapore, in 2005.  One of his research interests is to study first and second order methods to solve optimisation problems, such as, subgradient methods on convex/nonconvex optimisation problems and interior point methods on conic optimisation problems.  His other research interest is on supply chain optimisation, in particular, optimal inventory management.  He has published papers in Operations Research, SIAM Journal on Optimization, Mathematical Programming, European Journal of Operational Research, among others.   Previously, he held teaching and research positions in Hong Kong and Singapore, before joining the University of Portsmouth in 2014.

Key facts

In light of current Government legislation, we have postponed all seminars until 20 April. Should this date change, then we will update the seminars accordingly. For full University information regarding COVID-19 (coronavirus), please see our dedicated page