基于Wolfe线搜索的修正共轭梯度算法  

Modified conjugate methods based on the Wolfe line search

作　　者：林穗华 

机构地区：[1]广西民族师范学院数学与计算机科学学院,广西崇左532200

出　　处：《安徽大学学报（自然科学版）》2018年第2期47-53,共7页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(11261006);广西壮族自治区高等学校科学技术研究基金资助项目(ZD2014143);广西壮族自治区重点培育学科(应用数学)建设项目(桂教科研[2013]16);广西民族师范学院科研基金资助项目(2013RCGG002)

摘　　要：为解决大型无约束优化问题,设计新的修正参数公式,建立基于Wolfe线搜索的共轭梯度算法和谱共轭梯度算法,证明了新算法的下降性和全局收敛性.初步的数值实验表明算法是有效的.Motivated by solving large-scale unconstrained optimization problems, a new modified conjugate parameters were proposed. Based on the new conjugate parameters, a new conjugate gradient algorithm and a new spectral conjugate gradient algorithm with the Wolfe line search were presented, the descent property and the global convergence of the algorithms were proved. The preliminary numerical experiments showed that the algorithms were effective.

关 键 词：无约束优化 共轭梯度法 谱梯度法 WOLFE线搜索 全局收敛 

分 类 号：O221.2[理学—运筹学与控制论]