Cooperative Particle Swarm Optimization Decomposition Methods for Large-scale Optimization

Abstract

Many practical optimization problems can be classified as large-scale global optimization problems (LSOP). Various cooperative co-evolutionary (CC) algorithms have been proposed to combat the challenges of LSOPs. When CC algorithms are applied to large scale optimization problems, the effects of inter-connected variables, known as variable dependencies, can cause major performance degradation. Current literature provides different approaches to decomposing large-scale problems with variable dependencies during optimization using a wide range of base optimizers. In this paper, a cooperative particle swarm optimization (CPSO) algorithm is used as the base optimizer in a scalability study with a range of decomposition methods to determine ideal divide-and-conquer approaches when using a CPSO. Experimental results demonstrate that a variety of dynamic regrouping of variables, seen in the merging CPSO (MCPSO) and decomposition CPSO (DCPSO), as well varying total fitness evaluations per dimension, results in high-quality solutions when compared to six state-of-the-art decomposition approaches.