Differential Evolution with Cloud Model Based Self-adaptive Crossover Strategy

Efficiency of the evolutionary algorithms (EAs) is strongly dependent on the parameter setting. Differential evolution (DE) is well known as a simple and efficient evolutionary algorithm over continuous spaces. Of the three main parameters which control DE's behavior-population size NP, scaling factor F and crossover rate CR-CR is the most important. Despite the important role CR plays in effecting the performance of DE there are few studies that have examined the adaptive or self-adaptive strategies of CR. This paper is aim to investigate the influence of CR on the behavior of DE both from a theoretical and numerical point of view. In addition, a novel self-adaptive scheme for the evolution of CR based on a cloud model is proposed to enhance the convergence performance of DE. Experimental results confirm the superiority of the new scheme over several state-of-the-art self-adaptive DE versions.