The study on dynamic population size improvements for classical particle swarm optimization

In this work we presented two dynamic population size improvements for the classical PSO. EP-PSO started with a small number of particles and increased the number of particles dynamically by iteratively duplicating the updated particles. DP-PSO started with a large number of particles then reduced the number by dropping the worst performing half iteratively. Both EP-PSO and DP-PSO reduced the execution time by 60% on average compared to the classical PSO. EP-PSO fared quite badly when convergence rate and convergence ability to the global optimum was considered. On the other hand, DP-PSO performed reasonably well compared to the classical PSO but at a much faster convergence and execution speed.