论文部分内容阅读
系统分析现存多目标进化算法中分布度评价方法的特点和不足,提出一种基于最小生成树的可变邻域分布度评价方法,通过评价解集在“邻域”内的相对均匀程度,准确给出解集的分布结果,并部分解决现有方法不能对Pareto 最优面为非均匀分布的测试函数评价的问题.另外,给出一种解集映射方法,使其在少考虑一维信息同时,保持分布情况不变.实验结果证明该方法的可行性和有效性.
This paper systematically analyzes the characteristics and shortcomings of the existing evaluation methods for distributional degree in multi-objective evolutionary algorithms, and proposes a method for evaluating the distribution of variable neighborhoods based on minimum spanning tree. By evaluating the relative uniformity of solution sets in "neighborhood , And gives the distribution result of the solution set accurately, and partially solves the problem that the existing method can not evaluate the test function of the Pareto optimal surface as the non-uniform distribution.In addition, a solution mapping method is given, In the meantime, the dimension information keeps unchanged, and the experimental results prove the feasibility and effectiveness of the method.