Cooperative Coevolutionary NSGA-II with Linkage Measurement Minimization for Large-Scale Multi-objective Optimization

Rui Zhong, Masaharu Munetomo

Abstract: In this paper, we propose a novel decomposition method based on cooperative coevolution (CC) to deal with large-scale multi-objective optimization problems (LSMOPs) named Linkage Measurement Minimization (LMM), and after decomposition, NSGA-II is employed to optimize the subcomponents separately. CC is a mature and efficient framework for solving large-scale optimization problems (LSOPs), which decomposes LSOPs into multiple nonseparable subcomponents and solves them alternately based on a divide-and-conquer strategy. The essence of the successful implementation of the CC framework is the design of decomposition methods. However, in LSMOPs, variables in different objective functions may have different interactions, and the design of a proper decomposition method for LSMOPs is more difficult than for single objective optimization problems. Our proposed LMM can identify the relatively strong interactions and search the better decomposition iteratively. We evaluate our proposal on 21 benchmark functions of 500-D and 1000-D, and numerical experiments show that our proposal is quite competitive with the current popular decomposition methods.