一种新的混合共轭梯度算法

J4 ›› 2010, Vol. 45 ›› Issue (6): 81-85.

一种新的混合共轭梯度算法

程李晴^1,2, 石巧连²

1.郑州大学数学系，河南郑州 450001； 2.新乡医学院基础医学院, 河南新乡 453003

收稿日期:2009-09-07 出版日期:2010-06-16 发布日期:2010-06-17
作者简介:程李晴(1971-),女,讲师,硕士,研究方向:最优化理论和方法。Email:chengliqing@xxmu.edu.cn

A new hybrid conjugate gradient method

CHENG Li-qing^1,2, SHI Qiao-lian²

1. Department of Mathematics, Zhengzhou University, Zhengzhou 450001, Henan, China;
2. Basic Medical College, Xinxiang Medical University, Xinxiang 453003, Henan, China

Received:2009-09-07 Online:2010-06-16 Published:2010-06-17

摘要/Abstract

摘要：

给出了一种新的求解无约束优化问题的混合共轭梯度算法,该算法的搜索方向下降性不依赖于任何线搜索条件,并在Wolfe-Powell线搜索条件下证明了该算法具有全局收敛性,同时还给出了比较好的数值结果。

关键词: 无约束优化;混合共轭梯度法;Wolfe-Powell线搜索;全局收敛性

Abstract:

A new hybrid conjugate gradient formula for solving unconstrained optimization problem is proposed. The corresponding method can guarantee that the search directions are descent directions without any line search, and this algorithm with Wolfe-Powell line search is proved to be globally convergent. Preliminary numerical results show that the new hybrid conjugate gradient method is very efficient.

Key words: unconstrained optimization; hybrid conjugate gradient method; Wolfe-Powell line search; global convergence

程李晴1,2, 石巧连2. 一种新的混合共轭梯度算法[J]. J4, 2010, 45(6): 81-85.

CHENG Li-qing1,2, SHI Qiao-lian2. A new hybrid conjugate gradient method[J]. J4, 2010, 45(6): 81-85.

参考文献

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed