摘要:在多目标进化算法中，如何从后代候选集中选择最优解，显著地影响优化过程.当前，最优解的选择方式主要是基于实际目标值或者代理模型估计目标值.然而，这些选择方式往往是非常耗时或者存在精度差等问题，特别是对于一些实际的复杂优化问题.最近，一些研究人员开始利用有监督分类辅助后代选择，但是这些工作难以准备准确的正例和负例样本，或者存在耗时的参数调整等问题.为了解决这些问题，提出了一种新颖的融合分类与代理的混合个体选择机制，用于从后代候选集中选择最优解.在每一代优化中，首先利用分类器选择优良解；然后设计了一个轻量级的代理模型用于估计优良解的目标值；最后利用这些目标值对优良解进行排序，并选择最优解作为后代解.基于典型的多目标进化算法MOEA/D，利用混合个体选择机制设计了新的算法框架MOEA/D-CS.与当前流行的基于分解多目标进化算法比较，实验结果表明，所提出的算法取得了最好的性能.



Abstract:In multiobjective evolutionary algorithms, how to select the optimal solutions from the offspring candidate set significantly affects the optimization process. At present, the selection of the optimal solutions is largely based on the real objective values or surrogate model to estimate objective values. However, these selections are usually very time-consuming or of poor accuracy problems, especially for some real complex optimization problems. Recently, some researchers began to employ supervised classification to assist offspring selection, but these works are difficult to prepare the exact positive and negative samples or of time-consuming parameter adjustment problems. In order to solve these disadvantages, a novel hybrid individual selection mechanism is proposed through integrating classification and surrogate to select the optimal solutions from the offspring candidate set. Concretely, in each generation, the selection mechanism employs a classifier to select good solutions firstly; then, it designs a cheap surrogate model to estimate objective values of each good solution; finally, it sorts these good solutions according to objective values and selects the optimal solution as the offspring solution. Based on the typical multiobjective evolutionary algorithm MOEA/D, the hybrid individual selection mechanism is employed to design a new algorithm framework MOEA/D-CS. Compared with the current popular multiobjective evolutionary algorithms based on decomposition, experimental results show that the proposed algorithm obtains the best performance.



PDF全文下载地址:

http://jos.org.cn/jos/article/pdf/5602