Lévy Flight for Distribution-Based Discrete Particle Swarm Optimization

Abstract

Particle swarm optimization (PSO) is extended for discrete optimization and it has been shown to perform well. Some of the extended algorithms handle continuous parameters of probability distribution, which assume variable values of a candidate solution, instead of directly handling discrete variables. Such distribution-based discrete PSOs (DDPSOs) have drawback of their step size. We proposed a new sampling method to control the step size with Lévy distribution inspired by Lévy flight. Experimental results show the proposed method improves all the representative DDPSOs.