Adaptively choosing niching parameters in a PSO
Stefan Bird, Xiaodong Li · 2006
Niching techniques play an important role in evolutionary algorithms. Existing niching methods often require user-specified parameters, limiting their usefulness. This paper proposes a niching method for Particle Swarm Optimisation (PSO) where population statistics are used to adaptively determine the niching parameters during a run. The proposed niching method is compared to another niching based PSO, SPSO. Our results show that the proposed adaptive niching PSO can solve difficult multimodal functions more reliably and with fewer evaluations.