A diverse niche radii niching technique for multimodal function optimization

Fitness sharing based niching methods have been widely applied to multimodal optimization. However, they often suffer from the so-called niche radius problem as an improper selection of the niche radius may easily lead to suboptimal performance. To address such a problem, a diverse niche radii mechanism has been proposed in this paper. In the proposed algorithm, the radius of each niche will be individually adapted during the evolutionary process. This is achieved by devising and incorporating two operators, namely, niche combination and niche individual elimination, which are used to increase and decrease the radius of each niche, respectively. The proposed method has been evaluated on a set of standard functions. The results show that our method can significantly outperform related methods implemented for comparison.