一种新的混合谱共轭梯度法  

A New Hybrid Spectral Conjugate Gradient Method

作　　者：王森森 韩信 郑宗剑 WANG Sen-sen;HAN Xin;ZHENG Zong-jian(School of Mathematics and Information,Xinjiang Hetian College,Hetian 848000,China;School of Mathematics,Sichuan University of Arts and Science,Dazhou 635000,China;College of Electronics and Information Engineering,Southwest University,Chongqing 400715,China)

机构地区：[1]新疆和田学院数学与信息学院,新疆和田848000 [2]四川文理学院数学学院,四川达州635000 [3]西南大学电子信息工程学院,重庆400715

出　　处：《数学的实践与认识》2024年第7期150-157,共8页Mathematics in Practice and Theory

基　　金：政务数据安全达州市重点实验室一般项目(ZSAQ202311);达州市社科联重点研究基地数学与金融研究中心一般项目(SCMF202206);和田地区科技局项目(20220210)。

摘　　要：基于修正的共轭参数和谱共轭梯度法,提出了一种用于求解无约束优化问题的新混合谱共轭梯度法.提出的新算法在迭代过程中始终满足充分下降性.此外,新算法在标准Wolfe线搜索准则下是全局收敛的.与现存WYL共轭梯度法的改进版本进行数值仿真比较,结果表明所提方法具有更优异的计算性能.By virtue of the modified conjugate parameter and spectral conjugate gradient al-gorithm,a new hybrid spectral conjugate gradient method is proposed to tackle unconstrained optimization problems.The proposed method always the suficient descent property during the iteration process.Additionally,the new method has been proven to be globally convergent under the standard Wolfe line search condition.Compared with the existing improved version of the WYL conjugate gradient method through numerical simulation,the results evidence that the proposed method enjoys better computational performance.

关 键 词：无约束优化 混合谱共轭梯度法 标准Wolfe线搜索 全局收敛性 

分 类 号：O224[理学—运筹学与控制论]