AUSTRALASIAN CONFERENCE ON

ARTIFICIAL LIFE AND COMPUTATIONAL INTELLIGENCE (ACALCI 2016)

2-5 February 2016

Canberra, Australia

springer
LNAI

Kalyanmoy Deb

Kalynmoy Deb Biography: Kalyanmoy Deb is Koenig Endowed Chair Professor at Department of Electrical and Computer Engineering in Michigan State University, USA. Prior to his joining MSU, he was at Indian Institute of Technology (IIT) Kanpur. Prof. Deb's research interests are in evolutionary optimization and their application in optimization, modeling, and machine learning. He was awarded Infosys Prize, TWAS Prize in Engineering Sciences, CajAstur Mamdani Prize, Distinguished Alumni Award from IIT Kharagpur, Edgeworth-Pareto award, Bhatnagar Prize in Engineering Sciences, and Bessel Research award from Germany. He is fellow of IEEE, ASME, and three Indian science and engineering academies. He has published over 400 research papers with Google Scholar citation of 73,000 with h-index 90. He is in the editorial board on 20 major international journals. More information about his research contribution can be found from http://www.egr.msu.edu/~kdeb.

Title: A Convergence Measure Using Proximity to Optima in Single and Multi-Criterion Evolutionary Algorithms

Abstract:Theoretical optimality conditions, such as Karush-Kuhn-Tucker (KKT) conditions, are often used to assess whether a solution obtained by an evolutionary optimization (EO) algorithm is truly an optimum or not. However, recent studies have revealed that these conditions cannot provide a consistent evaluation of near-optimal solutions, hence making KKT conditions a `singular' event at the true optimum. In this talk, we argue that the KKT optimality conditions are not amenable to provide any neighborhood property of an optimal solution, thereby causing difficulties in using direct violation of KKT conditions as a measure of convergence property. Thereafter, we develop a KKT proximity metric that is able to provide a measure of "closeness" of a solution to the true optimal solution, which can be used as a termination condition for both single and multi-objective evolutionary optimization applications. Results using standard EO and EMO methods will be presented and their further use will be suggested.