The impact of population structure on particle swarm optimization: A network science perspective

Particle swarm optimization (PSO) is one of the most important swarm intelligence optimization algorithms due to its ease of implement and outstanding performance. As an information flow system, PSO is influenced by the population structure to a great extent. While previous works considered several classical structure, such as fully-connected and ring structures, here we systematically explore the impact of population structure, including scale-free and small-world networks that have been found in many real-world complex systems. In particular, we examine the influence of average degree, degree distribution and topological randomness of the networks underlying PSO. Our results are not only useful for developing more effective structures to improve the performance of PSO but also helpful in bridging the two fast-growing fields-network science and swarm intelligence.