Competitive Swarm Optimization with Dynamic Opposition-based Learning

Abstract

In order to enable the PSO to jump out of the local optima, we propose a Competitive Swarm Optimization with Dynamic Opposition-based learning (CSO-DOL). CSO-DOL contains two strategies: Competitive Learning and Opposition-based Learning. In each iteration, two randomly selected particles compete to get the winner and the loser. Then update the loser using opposition-based learning or competitive learning dynamically according to whether it falls into local optima to expand its search space. Compared with other state-of-art PSO variants on thirteen benchmark functions, the proposed algorithm can effectively help the particles jump out of the local optima on multimodal functions and has a faster convergence speed on simple unimodal functions.