Adaptive Radius Species Based Particle Swarm Optimization for Multimodal Optimization Problems

Multimodal optimization problem always has several peaks that are all optima of the problem. A promising approach to deal with such kind of problem should locate the peaks as many as possible (e.g., all the peaks) and should obtain high accuracy in each peak. The species-based particle swarm optimization (SPSO) divides the population into several subpopulations. Each subpopulation is gathered around a neighborhood best called species seed within the radius r, trying to locate different peaks. It does well in some low-dimensional multimodal optimization problems. However, the parameter r, which is associated with the efficiency and the accuracy of the algorithm, must be specified by the users. This makes SPSO very difficult for users to determine how much the parameter r should be. In this paper, a method of adaptively choosing radius r in SPSO is proposed, termed as adaptive SPSO (ASPSO). The experimental results show that the performance of ASPSO is more effective and accurate than standard SPSO in dealing with low-dimensional multimodal optimization problems.