Approximation Model Guided Selection for Evolutionary Multiobjective Optimization
Approximation Model Guided Selection for Evolutionary Multiobjective Optimization
复制标题
DOI：
10.1007/978-3-642-37140-0_31
复制
发表时间：
2013-03
期刊：
影响因子：
--
通讯作者：
Aimin Zhou;Qingfu Zhang;Guixu Zhang
中科院分区：
文献类型：
--
作者：
Aimin Zhou;Qingfu Zhang;Guixu Zhang
选择在多目标进化算法（MOEA）中起着关键作用。基于优势度的选择算子或基于指标的选择算子在当前的moea中被广泛使用。本文研究了另一种选择，即首先建立一个模型来逼近Pareto前沿，然后引导有希望的解的选择进入下一代。基于这一思想，本文提出了两种近似模型引导选择算子：一种是使用零阶模型逼近Pareto前沿，另一种是使用一阶模型逼近Pareto前沿。实验结果表明，新的AMS算子在一些测试实例上表现良好。
Selection plays a key role in amultiobjective evolutionary algorithm (MOEA). The dominance based selection operators or indicator based ones are widely used in most current MOEAs. This paper studies another kind of selection, in which a model is firstly built to approximate the Pareto front and then guides the selection of promising solutions into the next generation. Based on this idea, we propose twoapproximation model guided selection (AMS)operators in this paper: one uses a zero-order model to approximate the Pareto front, and the other uses a first-order model. The experimental results show that the new AMS operators performs well on some test instances.