Hybrid Weight Strategy for Particle Swarm Optimization

Abstract

Particle Swarm Optimization algorithm (PSO) is found to be an effective meta-heuristic swarm-based algorithm in solving modern time problems. Various improvements have been proposed in this algorithm in terms of internal computation, acceleration coefficients, stopping criteria, hybridization, velocity upgradation etc. The objective of this paper is to implement hybrid weights and, therefore, improve the quality of PSO algorithm. In the case of hybrid weights, we have combined two weights at a time. These weights are mixed in various but not in equal proportions and are tested against ten standard testing functions along with the pre-existing weights. By using this collection, we have analysed them on three parameters-mean, standard deviation, and minimum value achieved. Later on, after analysing the data, we found out that hybrid weights are an overall better option with respect to the pre-existing weights.