论文部分内容阅读
为有效利用决定空间中的信息、提高收敛速度与准确度,提出了基于决策空间划分模型的多目标进化算法.该算法将决策空间划分成多个子决策空间并在每个子决策空间内映射出一个超球体,运用某一多目标进化算法完成超球体内个体的1轮次进化,基于粒子群优化算法的粒子移动机制实现超球体间的信息共享、引导超球体质心向最优解集方向移动.对8个测试问题的实验结果表明:基于决策空间划分模型的多目标进化算法在收敛精度和收敛稳定性方面比FastPGA,MOCell,NSGA-Ⅱ和SPEA2算法表现出更好的性能.
In order to effectively utilize the information in decision space and improve the convergence speed and accuracy, a multi-objective evolutionary algorithm based on decision-making space partitioning model is proposed. The algorithm divides the decision space into multiple sub-decision spaces and maps one in each sub-decision space Hyperspheres, using a multi-objective evolutionary algorithm to complete a round of evolution of individuals in the hypersphere, the particle movement mechanism based on Particle Swarm Optimization algorithm to achieve information sharing between hyperspheres and guide the center of mass of the hypersphere to move toward the optimal solution set The experimental results on eight test problems show that the multi-objective evolutionary algorithm based on the decision-making space partitioning model performs better than the FastPGA, MOCell, NSGA-Ⅱ and SPEA2 algorithms in convergence accuracy and convergence stability.