Multipopulation evolution framework for multifactorial optimization

Abstract

Multifactorial 1 Optimization (MFO) has been attracting considerable attention in the evolutionary computation community. In this paper, we propose a general multi-population evolution framework (MPEF) for MFO, wherein each population has its own random mating probability (rmp) and is used for its own task. The benefits of using MPEF are twofold: 1) Various well-developed evolutionary algorithms (EAs) can be easily embedded into MPEF for solving the task(s) of MFO problems; 2) Different populations can implement different genetic material transfers. Moreover, for instantiation, we embed a powerful differential evolution algorithm, namely SHADE, into MPEF to form a multipopulation DE algorithm (MPEF-SHADE) for solving MFO problems. The experimental results on nine MFO benchmark problems show that MPEF-SHADE is significantly better than or at least competitive with other multifactorial evolution algorithms, such as MFEA, MFDE, MFPSO and AMA.