Computational Intelligence Method in Multi-Objective Optimization
Computational Intelligence Method in Multi-Objective Optimization
复制标题
DOI：
10.1109/sice.2006.315848
复制
发表时间：
2006-10
期刊：
影响因子：
--
通讯作者：
Y. Yun;Min Yoon;H. Nakayama
中科院分区：
文献类型：
--
作者：
Y. Yun;Min Yoon;H. Nakayama
决策问题可以表述为多目标优化问题，最终决策是从目标空间的Pareto最优解集合中做出的，该集合称为Pareto前沿。为了寻找Pareto边界，所谓的MOGA已被应用。另一方面，工程设计中的目标函数的形式不能以设计变量的形式明确给出。在这种情况下，目标函数的值可以通过一些分析来评估，这些分析通常是非常昂贵的。然而，现有的MOGA需要大量的功能评估产生帕累托最优解。因此，为了减少函数评估的数量，本文提出了一种混合技术的MOGA引入预测的目标函数的支持向量回归。通过数值算例，验证了该方法的有效性
Decision makings may be formulated as optimization problem with multiple objectives, and a final decision is made from the set of Pareto optimal solutions which is called as Pareto frontier in the objective space. For searching Pareto frontier, so-called MOGA has been applied. On the other hand, the forms of objective functions in engineering design cannot be given explicitly in terms of design variable. In this situation, the values of objective functions can be evaluated by some analyses, which are usually very expensive. However, existing MOGAs need a large number of function evaluations for generating Pareto optimal solutions. Therefore, in order to decrease the number of function evaluations, this paper proposes a hybrid technique of MOGA introducing a prediction of objective function by support vector regression. Through the numerical examples, the effectiveness of the proposed method will be shown