摘要
计算机试验异军突起,并因其经济性而越来越普遍地取代物理实验。Kriging模型作为计算机试验的元模型之一,因其使用简单、灵活被广泛地应用于各大领域。本文给出了基于Fiducial推断的Kriging模型选择方法,并与Lasso和Elastic Net惩罚下的选择方法相比较。数值模拟和实例分析表明Elastic Net惩罚下的选择方法优于Lasso,基于Fiducial推断的模型选择方法相较于Lasso和Elastic Net具有更高的拟合准确性和预测精度。
Computer experiments are becoming increasingly popular and surrogate for physical experiments because of their economy. Kriging model, as one of the meta models of computer experiments, is widely used in various fields because of its simplicity and flexibility. This paper studies a model se-lection method based on Fiducial inference for Kriging model, and compares with the selection methods under Lasso and Elastic Net penalties. Numerical simulation and case analysis show that the selection method based on Elastic Net penalty is superior to Lasso, and the model selection method based on Fiducial inference has higher fitting accuracy and prediction accuracy compared to Lasso and Elastic Net.
出处
《应用数学进展》
2024年第2期684-691,共8页
Advances in Applied Mathematics
