>> 自然科学版期刊 >> 2019年04期 >> 正文
一类基于Armijo线搜索的新的谱共轭梯度法
供稿: 景书杰;李亚敏;牛海峰 时间: 2019-07-04 次数:

作者 :景书杰李亚敏牛海峰

作者单位:河南理工大学数学与信息科学学院

摘要:为了构造具有更好收敛性的谱共轭梯度法,根据已有的共轭系数β~*_k和β■,构造了一个新的共轭系数β■,从而给出了一个新的谱共轭梯度法。经过选取适当的谱系数,保证新方法在每次迭代时总能产生充分下降的搜索方向。该性质具有既不依赖所使用的线搜索,又不依赖目标函数凸性的优点。利用Armijo线搜索,在一般假设条件下,给出了该方法全局收敛性的证明。

基金:国家自然科学基金资助项目(U1504104);

关键词:谱共轭梯度法;共轭系数;Armijo线搜索;全局收敛性;

DOI:10.16186/j.cnki.1673-9787.2019.4.23

分类号:O224

A new spectral conjugate gradient method based on Armijo line search

JING ShujieLI YaminNIU Haifeng

School of Mathematics and Information Science, Henan Polytechnic University

Abstract:In order to construct a new spectral conjugate gradient method with better convergence, a new conjugate coefficientβSNkkSΝwas constructed acccording to the existing conjugate coefficientsβ*kandβRMIL+kkRΜΙL+, thus a new specttral conjugate gradient method was given.The new method could always generate a sufficiently declining search drection in each iteration by selecting proper spectral coefficient.This property had an advantage that it depended neither on the line search which has been used nor on the convexity of the objective function.By using Armijo line search, the global convergence of this method was proved under general assumptions.

最近更新