Small-World Particle Swarm Optimization with Topology Adaptation

Abstract

Traditional particle swarm optimization (PSO) algorithms adopt completely regular network as topologies, which may encounter the problems of premature convergence and insufficient efficiency. In order to improve the performance of PSO, this paper proposes a novel topology based on small-world network. Each particle in the swarm interacts with its cohesive neighbors and by chance to communicate with some distant particles via small-world randomization. In order to improve search diversity, each dimension of the swarm is assigned with a specific network, and the particle is allowed to follow the historical information of different neighbors on different dimensions. Moreover, in the proposed small-world topology, the neighborhood size and the randomization probability are adaptively adjusted based on the convergence state of the swarm. By applying the topology adaptation mechanism, the particle swarm is able to balance its exploitation and exploration abilities during the search process. Experiments were conducted on a set of classical benchmark functions. The results verify the effectiveness and high efficiency of the proposed PSO algorithm with adaptive small-world topology when compared with some other PSO variants.