On Large Scale Evolutionary Optimization Using Simplex-Based Cooperative Coevolution Genetic Algorithm
Bo Yang, Hongfeng Xiao · 2009
Destruction of interdependencies of multivariable in decomposing hyper-high dimensional problems into single variable is generally the main reason that Conventional CC framework fails to optimize inseparable problems. An improved CC framework is proposed, which designs a basic optimizer that has better performance in high-dimension optimization. The optimizer is a simplex-based genetic algorithm (HD-simplex GA) that is composed by a fusion of the multi-direction searches of the Nelder-Mead simplex method and the evolution mechanism of steady GA. Based on above points, a simplex-based cooperative co-evolution genetic algorithm (Simplex CCGA) is presented. Extensive computational studies had been made to evaluate the performance of simplex CCGA in several benchmark functions with up to 500-1500 dimension. The results show that Simplex CCGA is more effective and efficient in the treatment of large scale optimization problems.