New evolutionary approach to the GCP: a premature convergence and an evolution process character

Abstract

This paper presents a new approach to the graph coloring problem (GCP) which utilizes information about conflict localization in a given coloring. In this context a partial fitness function (pff) and its usage to specialize genetic operators and phenotypic measure of diversity in population are described. Particular attention is given to the investigation of the influence of the population size and the usage of genetic operators on the character of the evolution, especially influence leading to a premature convergence in the evolution process. Experiments based on benchmark DIMACS graphs are presented.