A New Pre-initializing Strategy: Multi-Period Particle Swarm Optimization

A more efficient pre-initializing strategy of PSO algorithm: Multi-Period Particle Swarm Optimization (MP-PSO) is proposed. The process is divided into two periods: pre-initialization and post-optimization. The former is determined to find a better local solution to initialize the next period instead of standard uniform randomness. In order to explore further, adaptive escaping weight is adopted to avoid premature convergence during post-optimization. The results of benchmark test show that performance of MP-PSO is much more effective than that of standard PSO, especially in higher dimensional problems.