Cellular learning automata based bare bones PSO with maximum likelihood rotated mutations

Abstract

Abstract In this work, a new variant of multi-swarm bare bones PSO is presented which adaptively learns the promising alignments to re-orient its updating distributions. The idea is to maximize the likelihood of generating new particles along appropriate directions. As these alignment directions are unknown a priori, the proposed method employs a learning mechanism for adaptive learning of suitable alignments. The learning mechanism presented in this paper is based on cellular learning automata. Several alignment strategies are developed for each particle in the proposed method, and the cellular learning automata guides the particles toward the most promising directions by adjusting these strategies during the searching process. The new proposed algorithm is compared with a group of stochastic optimization approaches on the CEC2013 benchmark set. Experiential studies suggest the effectiveness of the proposed method in solving complex optimization problems.