A Grouping Cooperative Differential Evolution Algorithm for Solving Partially Separable Complex Optimization Problems

Differential evolution (DE) is a widely accepted optimization algorithm inspired by the mechanisms of biological evolution for complex optimization problems. In this paper, we put forward a co-evolutionary differential evolution (CDE) with a differential grouping (DG) mechanism to solve the complex optimization problems in which the variables are partially separable. In the CDE, a complex problem with coupled multivariable is decomposed by DG into certain independent sub-problems easy to be solved. Then polling, upper confidence bound (UCB), and random access are introduced to allocate search resources for multiple decoupled sub-problems, respectively. Finally, success-history-based parameter adaptation for differential evolution (SHADE) is adopted as a search engine to solve each sub-problem. The results of experiments on the CEC2017 show that CDE achieves a competitive search performance compared to other peer algorithms. This study suggests that the combination of DG strategy and polling method can effectively solve the optimization problem with partially separable variables.