Cooperative Co-evolution Algorithm with Problem Adaptive Variable Grouping for Large Scale Global Optimization
Fei Wei,Shugang Li,Jinfeng Xue
2018-01-01
Abstract:For large scale global optimization problems, evolutionary algorithms (briefly, EAs) will face a huge challenge, and their efficiency and effectiveness will be much reduced. To enhance their efficiency and effectiveness, in this paper, a problem adaptive variable grouping strategy (briefly, PAVG) is firstly proposed. In PAVG, we make the grouping directly via the expression of the objective function which usually consists of finite number of operations of four arithmetic operations “ + , − , 2 f , ÷ ” and composite operation of basic elementary functions. We classify these operations into two classes: one will result in non-separable variables, and the other will result in separable variables. In this way, the variables can be grouped into several non-interacting subcomponents, while the variables in each subcomponent are interactive. Then, combining with PAVG, a novel algorithm, called cooperative co-evolution algorithm with problem adaptive variable grouping strategy (briefly, CCPA), is designed, and evolution can be conducted in these subcomponents separately. In this way, a large-scale problem can be decomposed into several small-scale problems and this makes the problem solving much easier. To further enhance the efficiency of CCPA, a new local search scheme is designed, and when a good solution is found in the evolution process (e.g., by crossover and mutation), it will be further improved by the local search scheme. To verify the efficiency of CCPA, the simulations are made on the standard benchmark suites of CEC’2010 and CEC’2013, and CCPA is compared with several well performed algorithms. The results indicate that the proposed algorithm CCPA is more efficient and effective.