Particle Swarm Optimization Based on Self-organizing Topology Driven by Fitness

To explore the relation between topologic characteristics of dynamic network and performance of the particle swarm optimization (PSO) algorithm, the population of PSO is viewed as a network where each particle is represented as a node and the network structure changes dynamically as the fitness of particles varies. Moreover, in this paper, the structural changes involve adding and removing the links but the network size remains the same. Then, two kinds of simulations are conducted. The results from one kind focusing on PSO show that the dynamic network is capable of balancing exploration and exploitation so that the performance of PSO can be improved as long as the weight θ is selected properly. In addition, the results from other kind concerning on topologic characteristics of dynamical network indicate the impact of network structure on algorithm behavior and the law of network evolution.