Multiple crossover per couple and fitness proportional couple selection in genetic algorithms

Abstract

Contrasting with conventional approaches to crossover, Multiple Crossover Per Couple (MCPC) is an alternative, recently proposed [1], approach under which more than one crossover operation for each mating pair is allowed. In genetic algorithms, Proportional Selection (PS) is a popular method to select individuals for mating based on their fitness values. The Fitness Proportional Couple Selection (FPCS) approach, is a new selection method which creates an intermediate population of couples from where, subsequently, couples are selected for crossing-over based on couple fitness. This paper proposes the combined use of MCPC and FPCS. Outstanding performance was achieved by joining both methods when optimising hard testing multimodal and unimodal functions. Some of these results and their comparison against results from conventional approaches are shown.