Improving the Particle Swarm Optimizer by Function “Stretching”

Abstract

In this paper a new technique, named Function “Stretching”, for thealleviation of the local minima problem is proposed. The main featureof this technique is the usage of a two–stage transformation of theobjective function to eliminate local minima, while preserving theglobal ones.Experiments indicate that combined with the Particle Swarm Optimizermethod, the new algorithm is capable of escaping from local minimaandeffectively locate the global ones. Our experience is that the modifiedalgorithm behaves predictably and reliably and the results were quitesatisfactory. The function “Stretching” technique provides stableconvergence and thus a better probability of success to the methodwithwhich it is combined.