Hierarchical Particle Swarm Optimization Algorithm for Multimodal Function Optimization

In this paper, we propose a Hierarchical Particle Swarm Optimization (HPSO) algorithm model for multimodal function optimization. All particles will be classified into several groups, these groups operated at two levels: one level is to find location of global optima clusters with its particles; the other is to exactly distinguish multiple global optima in clusters. By this mechanism, algorithm can explore the search space at different level of scales simultaneously; all the global optima can be detected effectively. In addition, the algorithm is easy to be controlled; only two parameters are needed to be predefined before the running of algorithm and algorithm’s performance is not sensitive to the influence caused by these two parameters. We tested our algorithm on several benchmark functions and compared it with some well-known methods for these functions. Experimental results showed that our algorithm could achieve an overall good performance. We also analyzed the effect of the parameters for our algorithm performance at the end of our paper.