Particle Swarms with dynamic ring topology
Yuxuan Wang, Qiao-Liang Xiang · 2008
Particle Swarm Optimizer (PSO) is a recently proposed population-based evolutionary algorithm, which exhibits good performance in many fields, and now it’s becoming more and more popular due to its strong global optimization capability and simple implementation. To achieve better performance, some variants investigated the utilization of different topologies in PSO. However, particles are only “conceptually” connected in the topology, and the neighborhoods of a certain particle never change (i.e. the neighborhood structure is fixed). In this paper, we propose a dynamically changing ring topology, in which particles are connected unidirectionally with respect to their personal best fitness. Meanwhile, two strategies, namely the “Learn From Far and Better Ones” strategy and the “Centroid of Mass” strategy are used to enable certain particle to communicate with its neighbors. Experimental results on six benchmarks functions validate the effectiveness of the proposed algorithm.