Problem Structure and the Behavior of Evolutionary Algorithms

Dr Andrew Sutton

School of Computer Science, University of Adelaide, Australia

Date and time: 11.30 - 12.30, Thursday 2nd February, 2012

Venue: 14.08.10 (Building 14, Level 8, Intelligent Systems Area)

Abstract:

Evolutionary algorithms are a class of nature-inspired randomized search heuristics that are often tasked to solve hard combinatorial optimization problems. Despite their popularity, very little is rigorously understood about when (and how) they work well and when (and why) they fail.

In this talk, I will give a broad overview of research that aims to understand how the structure of different combinatorial problems affects the behavior of evolutionary algorithms. I will outline recent results from the analysis of fitness landscapes. I will also discuss some recent developments in parameterized analysis for EAs which aims to express the runtime of EAs in terms of a parameter that captures the source of exponential complexity in the structure of the problem.

About the speaker:

Andrew Sutton completed his Ph.D. at Colorado State University investigating properties of combinatorial fitness landscapes. He currently holds a postdoctoral research fellowship at the University of Adelaide studying theory of evolutionary algorithms under the supervision of Dr. Frank Neumann.


Seminar Organisation

Seminars are free and open to the general public. No booking is necessary. If you are interested in giving a presentation in this seminar series, or to make suggestions for speakers, please contact Xiaodong Li, the seminar co-ordinator.