Particle Swarm Optimization with Dynamic Dimension Crossover for High Dimensional Problems

Previous work presented some modified approaches based particle swarm optimization (PSO) to solve complex optimization problems. Preliminary results demonstrated that PSO with crossover (CPSO) constituted a promising approach to solve some optimization problems. However how to optimize high dimensional problem with crossover became challenging. In this paper, a modified PSO with dimension crossover is proposed. First we analyze the cause of hardly optimizing the high dimensional problem, and then design one dynamic dimension crossover PSO (DDC-PSO) to cope with high dimensional problems. Theoretical analysis is also presented to show why the modified algorithm can be effective. Finally DDC-PSO is tested on five benchmark optimization problems and the results show a superior performance compared to the standard PSO and CPSO.