A New Three-Level Recursive Differential Grouping Method for Large-Scale Optimization

Abstract

Cooperative co-evolution (CC) is an evolutionary framework that breaks down a complex problem into several sub-problems and then solves them separately. It is particularly effective in dealing with large-scale optimization problems. Recently, a three-level recursive difference decomposition method has been proposed. This method can divide one set into three equal parts according to the interaction between sets, and then detect the interaction between the subsets and another set. In this paper, a novel three-level recursive differential decomposition method is proposed. Based on the former method, a reference to historical interactive information is introduced, which significantly improves the decomposition efficiency. This decomposition method performs well in CEC2010 and CEC2013 test functions, saves more calculation than other differential grouping methods, and has good competitiveness in the two CC frameworks.