Hybrid evolutionary algorithms design based on their advantages

Abstract

The search direction and the search step size are two important factors which affect the performance of algorithms. In this paper, we combine Particle Swarm Optimization (PSO) with EP to form two new algorithms namely PSOEP and SAVPSO. The basic idea is to introduce the search direction to the mutation operator of EP and use lognormal self-adaptive strategy to control the velocity of PSO to guide the individual at a faster convergence rate. All of these algorithms are compared to each other with respect to the similarities and differences of their basic components, as well as their performances on seven benchmark problems. Our experimental results show that PSOEP performs much better than all other version of EPs, and SAVPSO performs much better than PSO for the benchmark functions. © 2010 Springer-Verlag Berlin Heidelberg.