【摘要】 代理模型辅助的多目标进化算法广泛用于解决计算费时的多目标优化问题，然而现有的大部分建模方法都是为了嵌入到特定算法而设计的，适应于其他算法的能力并不强，为了能够依据数据特征自适应的建立模型，提出了一种基于自适应模型选择的建模方法。该方法的主要思想为：依据每个目标函数的样本特征，自适应的选择样本建立全局模型或者局部模型。为了验证所提出建模的方法的有效性，将提出的建模方法应用于基于高斯过程辅助的双存档费时多目标优化算法(KAT2)和基于高斯过程辅助的参考向量引导的费时多目标优化算法(K-RVEA),并且在DTLZ测试函数进行测试。通过实验证明，提出的建模方法可以有效的解决费时多目标优化问题。更多还原

【Abstract】 Surrogate assisted multi-objective evolutionary algorithm are widely used to solve time-consuming multi-objective optimization problems.However, most existing modeling methods are designed to be embedded into specific algorithms, and the ability to adapt to other algorithms is not strong.In order to build models adaptively according to data characteristics, a modeling method based on adaptive model selection is proposed.The main idea of this method is to adaptively select samples to establish global model or local model according to the sample characteristics of each objective function.In order to verify the effectiveness of the proposed modeling method, the proposed modeling method is applied to the double archiving time-consuming multi-objective optimization algorithm based on Gaussian process assistance(KAT2) and the time-consuming multi-objective optimization algorithm guided by reference vector based on Gaussian process assistance(K-RVEA),and tested in dtlz test function.Experiments show that the proposed modeling method can effectively solve the time-consuming multi-objective optimization problem.更多还原