A novel particle swarm optimization algorithm with stochastic focusing search for real-parameter optimization

Abstract

Basic particle swarm optimization (PSO) algorithms are susceptible to being trapped into local optimum and premature convergence. A novel stochastic focusing search (SFS) based PSO algorithm with adaptively dynamic neighborhoods topology and subpopulation strategy is proposed. SFS is based on simulating the act of human randomized searching behaviors by using the adaptively dynamic neighborhoods topology. With subpopulation strategy, SFS can improve the global searching ability, keeping the diversity and escaping local extremum. The algorithm?s performance is studied by using a challenging set of typically complex functions with comparison of differential evolution (DE) and three modified PSO algorithms. The simulation results show that SFS is competitive to solve most parts of the benchmark problems and will become a promising candidate of search algorithms especially when the existing algorithms have difficulties in solving some problems.