A Simple and Fast Particle Swarm Optimization
Hui Wang, Zhijian Wu, Sanyou Zeng, Dazhi Jiang, Yong Liu, Jing Wang and Xianqiang Yang
Particle Swarm Optimization (PSO) has shown its good performance on well-known numerical function problems. However, on some multimodal functions the PSO easily suffers from premature convergence because of the rapid decline in diversity. Some diversity-guided PSO algorithms have been proposed to maintain diversity, while these techniques cost much computation time on the calculation of diversity. In this paper, a simple and fast PSO (hybrid PSO, namely HPSO) is proposed, which indirectly maintains the diversity of swarm but not compute it. Experimental studies on 16 well-known benchmark functions show that the HPSO not only obtains better performance than the standard PSO and other two diversity guided PSO algorithms, but almost cost the same computation time with the standard PSO. In addition, a comprehensive set of experiments including the average computation time, the effects of crossover rate (CR) on the performance of HPSO, the successful rate of the elitist selection and the effects of CR on the diversity are empirically verified.
Keywords: Particle swarm optimization (PSO), diversity-guided, function optimization, diversity measure, elitist selection, computation time.