自适应偏好半径划分区域的多目标进化方法

发布时间：2018-07-13 10:29

【摘要】：偏好多目标进化算法是一类帮助决策者找到感兴趣的Pareto最优解的算法.目前,在以参考点位置作为偏好信息载体的偏好多目标进化算法中,不合适的参考点位置往往会严重影响算法的收敛性能,偏好区域的大小难以控制,在高维问题上效果较差.针对以上问题,通过计算基于种群的自适应偏好半径,利用自适应偏好半径构造一种新的偏好关系模型,通过对偏好区域进行划分,提出基于偏好区域划分的偏好多目标进化算法.将所提算法与4种常用的以参考点为偏好信息载体的多目标进化算法g-NSGA-II、r-NSGA-II、角度偏好算法、MOEA/D-PRE进行对比实验,结果表明,所提算法具有较好的收敛性能和分布性能,决策者可以控制偏好区域大小,在高维问题上也具有较好的收敛效果.
[Abstract]:Preference multiobjective evolutionary algorithm is a kind of algorithm that helps decision makers to find the Pareto optimal solution of interest. At present, in the preference multi-objective evolutionary algorithm with reference point position as the carrier of preference information, inappropriate reference point location often seriously affects the convergence performance of the algorithm, the size of preference region is difficult to control, and the effect on high-dimensional problem is poor. To solve the above problems, a new preference relation model is constructed by calculating the adaptive preference radius based on population, and the preference region is divided. A preference multi-objective evolutionary algorithm based on preference region partition is proposed. The proposed algorithm is compared with four commonly used multi-objective evolutionary algorithms, g-NSGA-IIr-NSGA-IIE, and angle preference algorithm, MOEA-D-PRE. The results show that the proposed algorithm has better convergence and distribution performance. Decision makers can control the size of preference regions and have good convergence effect on high dimensional problems.
【作者单位】：智能计算与信息处理教育部重点实验室(湘潭大学);智能信息处理与应用湖南省重点实验室(衡阳师范学院);Computer
【基金】：国家自然科学基金(61502408,61673331,61379062,61403326) 湖南省教育厅重点项目(17A212) 赛尔网络创新项目(NGII20150302) 湖南省自然科学基金(14JJ2072,2017JJ4001) 湖南省科技计划(2016TP1020)~~
【分类号】：TP18

【相似文献】