Parameter control methods for selection operators in genetic algorithms

Abstract

Parameter control is still one of the main challenges in evolutionary computation. This paper is concerned with controlling selection operators on-the-fly. We perform an experimental comparison of such methods on three groups of test functions and conclude that varying selection pressure during a GA run often yields performance benefits, and therefore is a recommended option for designers and users of evolutionary algorithms. © 2008 Springer-Verlag Berlin Heidelberg.