On the Performance of Particle Swarm Optimization Algorithms in Solving Cheap Problems

Abstract

Eight variants of the Particle Swarm Optimization (PSO) algorithm are discussed and experimentally compared among each other. The chosen PSO variants reflect recent research directions on PSO, namely parameter tuning, neighborhood topology, and learning strategies. The Comparing Continuous Optimizers (COCO) methodology was adopted in comparing these variants on the noiseless BBOB test bed. Based on the results, we provide useful insights regarding PSO variants' relative efficiency and effectiveness under a cheap budget of function evaluations, and draw suggestions about which variant should be used depending on what we know about our optimization problem in terms of evaluation budget, dimensionality, and function structure. Furthermore, we propose possible future research directions addressing the limitations of latest PSO variants. We hope this paper would mark a milestone in assessing the state-of-the-art PSO algorithms, and become a reference for swarm intelligence community regarding this matter.