›› 2015, Vol. 28 ›› Issue (5): 32-.

• 论文 • 上一篇    下一篇

一种基于梯度法的Kriging参数优化算法

李永   

  1. (西安电子科技大学 数学与统计学院,陕西 西安 710126)
  • 出版日期:2015-05-15 发布日期:2015-05-19
  • 作者简介:李永(1988—),男,硕士研究生。研究方向:基于Kriging模型的全局优化算法。Email:leheliyong@126.com

A Kriging Parameters Optimization Algorithm Based on Gradient Method

LI Yong   

  1. (School of Mathematics and Statistics,Xidian University,Xi'an 710126,China)
  • Online:2015-05-15 Published:2015-05-19

摘要:

相关函数参数的确定是Kriging模型构造的关键,针对传统模式搜索方法不够精确的缺点,文中提出用投影梯度法求解Kriging模型中空间相关函数参数θ的算法,对目标函数求解关于θ的一阶梯度,同时在单步求解中保持回归系数β不变。数值实验结果表明,这种方法能够得到更为精确的结果。

关键词: Kriging模型, 相关函数参数, 投影梯度法

Abstract:

The key to the construction of the Kriging model is to determine the parameter of the correlation function.To solve the problem of low accuracy of the traditional pattern search method,we propose a method using the projection gradient method to solve the Kriging algorithm of spatial correlation function parameters.For the target function,we also solve the first order gradient about  θ with the value of β  kept constant in stepbystep solution.The outcome of numerical experiments shows that we can obtain more accurate solution by this method.

Key words: Kriging model;correlation function parameters;projection gradient method

中图分类号: 

  • TP301.6