惩罚策略辅助的快速非支配排序遗传算法Ⅱ研究

Research on fast non-dominated sorting genetic algorithmⅡassisted by penalty strategy

快速非支配排序遗传算法Ⅱ(NSGA-Ⅱ)是一种典型的多目标优化算法。然而,其采用的锦标赛选择策略存在着重复选择较优个体的缺陷,由此导致产生的后代分布性较差以及算法整体性能下降。为解决锦标赛选择策略的缺陷,提出惩罚策略辅助的锦标赛选择策略。具体而言,在通过惩罚策略辅助的锦标赛选择策略选择交叉个体时,在下一轮中每个被选择个体的优先级自动降低,以此降低较优个体被选中的概率。将基于惩罚策略辅助的锦标赛选择策略融入NSGA-Ⅱ算法,提出了惩罚策略辅助的快速非支配排序遗传算法Ⅱ。在ZDT和DTLZ测试集上与多个算法进行对比,结果表明:所提策略有效地解决了原锦标赛选择策略的缺陷,说明了改进策略的有效性。

Fast non-dominated sorting genetic algorithmⅡ(NSGA-Ⅱ)is a typical multi-objective optimization method,the tournament selection strategy employed in NSGA-Ⅱhas the drawback of repeatedly selecting solutions with better performance,resulting in the poor distribution of offspring solutions and the reduction of the overall performance.To tackle the drawback of the tournament selection strategy,this paper proposes a penalty strategy assisted tournament selection strategy.To be specific,when performing the penalty strategy assisted tournament selection,the priorities of selected solutions are lowered adaptively to reduce the probability that better solutions are selected in the next round.Further,this paper incorporates the penalty strategy assisted tournament selection strategy into NSGA-Ⅱand proposes a fast Non-dominated Sorting Genetic AlgorithmⅡassisted by Penalty Strategy.Experimental results on ZDT and DTLZ test suites against multiple comparison methods demonstrate that the proposed strategy indeed effectively overcomes the drawback of the original tournament selection strategy,illustrating the effectiveness of the proposed strategy.