On the influence of selection schemes on the genetic diversity in genetic algorithms

Abstract

This paper discusses some aspects of the general convergence behavior of genetic algorithms. Careful attention is given to how different selection strategies influence the progress of genetic diversity in populations. For being able to observe genetic diversity over time measures are introduced for estimating pairwise similarities as well as similarities among populations; these measures allow different perspectives to the similarity distribution of a genetic algorithm's population during its execution. The similarity distribution of populations is illustrated exemplarily on the basis of some routing problem instances. © 2009 Springer-Verlag Berlin Heidelberg.