Collective dynamics of particle swarm optimization: A network science perspective

Particle swarm optimization (PSO) is a cornerstone of evolutionary computation, yet its population dynamics and topological properties remain poorly understood beyond traditional stability analysis. This study presents the first network science-based investigation of PSO’s intrinsic topology, demonstrating that its network structure inherently exhibits small-world architecture and heavy-tailed degree distributions. Through systematic analysis of 13 benchmark functions – including 7 unimodal and 6 multimodal problems – we construct population communication networks where nodes represent particles and edges denote the interaction between individuals. This interdisciplinary lens provides a promising theoretical framework for analyzing evolutionary computation methods.