An Evolutionary Dynamic Population Size PSO Implementation

Particle Swarm Optimization (PSO) is a heuristic search method for the exploration of solution spaces of complex optimization problems. The heuristic suffers from relatively long execution times as the update step needs to be repeated many thousands of iterations to converge the swarm on the global optimum. In this work, we explore two dynamic population size improvements for classical PSO with the aim of reducing execution time. Expanding Population PSO (EP-PSO) starts with a small number of particles and iteratively increases the swarm size. Diminishing Population PSO (DP-PSO) starts with a large number of particles and iteratively reduces the swarm size. Simulation results show that both improvements produce almost 60% reduction in the execution time as compared to the classical PSO. However, the results show that EP-PSO fares quite badly when the ability to converge to the global optimum is concerned. DP-PSO performs reasonably compared to the classical PSO but at much faster convergence and execution speeds. Clearly, DP-PSO shows a lot of promise as an enhancement for the classical PSO.