A Variant of Unified Bare Bone Particle Swarm Optimizer

Abstract

The simplicity of bare bone particle swarm optimization (BPSO) is attractive since no parameters tuning is required. Nevertheless, it also encounters the issue of premature convergence. To remedy this problem, by integrated global model and local model search strategies, a unified bare bone particle swarm optimization (UBPSO) is appeared in recently where the weightings of global and local search strategies may be constant or random varying. In this paper, a variant of UBPSO is proposed that stresses on global exploration ability in early stages and turns to local exploitation in later stages for searching optimal solution. Numerical results reveal that this variant is competitive to UBPSO and performs better than BPSO and PSO in most of the tested benchmark functions.