Extended social learning guided particle swarm optimization

摘要

In this paper social learning in particle swarm optimization is extended. A particle not only exchanges information with the best in its group, but also learns from an ensemble guide which combines some previous best positions of the particles using ensemble learning technique. In addition, a whole swarm is divided into several parts and in each sub swarm, a particle also learns from another sub swarm's best particle. Based on these, an improved algorithm, named extended social learning guided particle swarm optimization (EGPSO), is proposed. Ensemble learning can help providing a more accurate global guide and learning from other groups can help increasing diversity. This algorithm is compared with standard PSO and some other improved PSO algorithms to illustrate how EGPSO can benefit from these strategies.