Parallel Coevolution of Quantum-Behaved Particle Swarm Optimization for High-Dimensional Problems

Quantum-behaved particle swarm optimization (QPSO) has successfully been applied to unimodal and multimodal optimization problems. However, with the emergence and popularity of big data and deep machine learning, QPSO encounters limitations with high dimensional problems. In this paper, a parallel coevolution framework of QPSO (PC_QPSO) is designed, in which an improved differential grouping method is used to decompose the high dimensional problems into several sub-problems. These sub-problems are optimized independently with occasional communication. Each sub-population is evaluated with context vector, which is constituted by the global best solutions in each sub-problem. The numerical experimental results show that PC_QPSO with differential grouping strategy is able to improve the solution quality without breaking the relationship between interacted variables.