Wolfe线搜下改进的FR型谱共轭梯度法

打开文本图片集
摘 要:谱共轭梯度法作为经典共轭梯度法的推广,它是求解大规模无约束优化问题的有效方法之一.基于标准Wolfe线搜索准则和充分下降性条件,提出了一种具有充分下降性质的FR型谱共轭梯度法.在温和的假设条件下,该算法具有全局收敛性.最后,将新算法与现存的修正FR型谱共轭梯度法进行比较,数值结果表明提出的算法是极其有效的.
关键词:无约束优化;谱共轭梯度法;充分下降性;标准Wolfe线搜索准则;全局收敛性
中图分类号:O221.2 文献标识码:A 文章编号:1009-3583(2024)-0080-05
FR Type Spectral Conjugate Gradient Method Improved
by Wolfe Line-search
(WANG Sen-sen1, HAN Xin2*, WU Xiang-biao3
(1.School of Mathematics and Information Science, Xinjiang Hetian College, Hetian 848000, China; 2. School of Mathematics, Sichuan University of Arts and Sciences, Dazhou 635000, China; 3. School of Mathematics, Zunyi Normal University, Zunyi 563006, China)
Abstract: The spectral conjugate gradient method, as an extension of the classical conjugate gradient method, is one of the effective methods for solving large-scale unconstrained optimization problems. Based on the standard Wolfe line search criterion and sufficient descent condition, a FR type spectral conjugate gradient method with sufficient descent property is proposed. Under mild assumptions, the algorithm has global convergence. Finally, the new algorithm is compared with the existing modified FR type spectral conjugate gradient method, and numerical results show that the proposed algorithm is extremely effective.
Keywords: unconstrained optimization; spectral conjugate gradient method; sufficient degradability; standard Wolfe line search criteria; global convergence
共轭梯度法作为一种优化方法,凭借其算法结构简单、易于编程和存储需求少等特点,常被用于解决航空航天、大气模拟、石油勘探、信号恢复等工程应用领域遇到的大规模优化问题[1-4].共轭梯度法 (Conjugate Gradient Method),简称CG法,最早是由Stiefel和Hestenes在求解非线性方程组时提出的算法,1962年Reeves和Fletcher将此方法应用于解决非线性优化问题。(剩余3570字)