Performance improvement of particle swarm optimization for high-dimensional function optimization

Particle swarm optimization (PSO) is a kind of population-based search methods, that is inspired by social behavior observed in nature, such as flocks of irds and schools of fish. PSO has been receiving attentions, since it has a powerful search ability in function optimization problems, and several improvement has been studied to apply PSO to the multimodal function optimization and optimization in the dynamic environments. The purpose of this paper is to improve PSO performance deteriorated by the degeneracy of particle velocities, in case of high-dimensional optimization problems. We propose a novel PSO model, called the Rotated Particle Swarm (RPS), by introducing the coordinate conversion. The numerical simulation results show that the proposed RPS is effective in optimizing high-dimensional functions.