Why simple population restart does not work in PSO

Abstract

In this study, we present an in-deep study on the poor ability of Particle Swarm Optimization (PSO) to recover from a premature convergence by simple population restart. Despite that, population restart is a popular component of many adaptive metaheuristic methods it has rarely been successfully utilized in PSO. We describe the methodology for basic population restart and observe the population diversity and performance change after the population is restarted.