Multilevel cooperative coevolution for large scale optimization

Zhenyu Yang, Ke Tang, Xin Yao

Abstract: In this paper, we propose a multilevel cooperative coevolution (MLCC) framework for large scale optimization problems. The motivation is to improve our previous work on grouping based cooperative coevolution (EACC-G), which has a hard-to-determine parameter, group size, in tackling problem decomposition. The problem decomposer takes group size as parameter to divide the objective vector into low dimensional subcomponents with a random grouping strategy. In the MLCC, a set of problem decomposers is constructed based on the random grouping strategy with different group sizes. The evolution process is divided into a number of cycles, and at the start of each cycle MLCC uses a self-adapted mechanism to select a decomposer according to its historical performance. Since different group sizes capture different interaction levels between the original objective variables, MLCC is able to self-adapt among different levels. The efficacy of the proposed MLCC is evaluated on the set of benchmark functions provided by CECpsila2008 special session.