Synergy of Differential Evolution and Particle Swarm Optimization

Abstract

A novel hybrid swarm intelligent algorithm of DE and PSO is proposed in this paper based on a tri-breakup concept in the population. The algorithm thus design is named as DE-PSO-DE, shortly DPD algorithm. By this proposed mechanism, all individuals in the population are portioned into three group’s namely inferior, mid and superior group, based on their fitness values. DE is employed to the inferior and superior group whereas; PSO is used in the mid group. Initially the suitable mutation operators for both DEs (used in DPD) are investigated. Later, top 4 DPDs (Viz. DPDs those use best 4 combinations of mutation strategies in DE) are chosen for further study. DPD combines the global search ability of DE and the local search ability of PSO. The process of hybridization offsets the weaknesses of each other. In addition, two strategies namely Elitism and Non-redundant search have been incorporated in DPD cycle. The supremacy of DPD is realized over a set of typical unconstrained benchmark suite problems. Lastly, all top 4 DPDs are used in solving three real life problems. The numerical and graphical results confirm that the proposed DPD yields more accurate values with faster convergence rate. Finally, a particular DPD is recommended as the best amongst all DPDs to solve unconstrained global optimization problems.