Towards a More Efficient Multi-Objective Particle Swarm Optimizer
Luis V. Santana-Quintero (CINVESTAV-IPN, Evolutionary Computation Group (EVOCINV), Mexico), Noel Ramírez-Santiago (CINVESTAV-IPN, Evolutionary Computation Group (EVOCINV), Mexico) and Carlos A. Coello Coello (CINVESTAV-IPN, Evolutionary Computation Group (EVOCINV), Mexico)
Copyright: © 2008
This chapter presents a hybrid between a particle swarm optimization (PSO) approach and scatter search. The main motivation for developing this approach is to combine the high convergence rate of the PSO algorithm with a local search approach based on scatter search, in order to have the main advantages of these two types of techniques. We propose a new leader selection scheme for PSO, which aims to accelerate convergence by increasing the selection pressure. However, this higher selection pressure reduces diversity. To alleviate that, scatter search is adopted after applying PSO, in order to spread the solutions previously obtained, so that a better distribution along the Pareto front is achieved. The proposed approach can produce reasonably good approximations of multi-objective problems of high dimensionality, performing only 4,000 fitness function evaluations. Test problems taken from the specialized literature are adopted to validate the proposed hybrid approach. Results are compared with respect to the NSGA-II, which is an approach representative of the state-of-the-art in the area.