一种基于云模型的多目标进化算法

在多目标进化算法的基础上,提出了一种基于云模型的多目标进化算法(CMOEA).算法设计了一种新的变异算子来自适应地调整变异概率,使得算法具有良好的局部搜索能力.算法采用小生境技术,其半径按X条件云发生器非线性动态地调整以便于保持解的多样性,同时动态计算个体的拥挤距离并采用云模型参数来估计个体的拥挤度,逐个删除种群中超出的非劣解以保持解的分布性.将该算法用于多目标0/1背包问题来测试CMOEA的性能,并与目前最流行且有效的多目标进化算法NSGA-II及SPEA2进行了比较.结果表明,CMOEA具有良好的搜索性能,并能很好地维持种群的多样性,快速收敛到Pareto前沿,所获得的Pareto最优解集具有更好的收敛性与分布性.

A Multi-Objective Evolutionary Algorithm Based on Cloud Model

A cloud model-based multi-objective evolutionary algorithm(CMOEA) is proposed based on the multiobjective evolutionary algorithm.In CMOEA,a new mutation operator that adaptively adjusts the mutation probability is designed to guarantee the good local searching ability.To maintain the diversity of solutions,the niche technology is exploited,where the niche radius is dynamically adjusted according to the X conditions cloud generator.Meanwhile,the dynamic calculation of crowding distance for individuals and the estimation of the individual congestion intensity by the cloud model are conducted at the same time,which is then followed by the eliminating process that removes the excess population one by one to keep non-inferior solutions for distribution.Finally,the multi-objective 0/1 knapsack problem is employed to test the performance of CMOEA.Experimental results indicate that compared with the currently most effective multiobjective evolutionary algorithms(NSGA-II and SPEA2),CMOEA has a better performance in searching and population diversity.In addition,fast convergence to the Pareto front is also achieved and the resulting set of Pareto optimal solutions has superior convergence and distribution.