摘要
提出了一类WFR型谱共轭梯度法,并且该算法在任何线搜索下都具有充分下降性.在标准Wolfe线搜索下,证明了新算法具有全局收敛性.数值实验结果表明新算法优于VFR法.
In this paper,a spectral conjugate gradient WFR method is put forward,which always possesses the sufficient descent property with any line search.It is proved under the standard Wolfe line search that the new spectral conjugate gradient method possesses global convergence.A series of numerical tests indicate that the new algorithm is superior to the VFR method.
出处
《西南大学学报(自然科学版)》
CAS
CSCD
北大核心
2018年第2期49-55,共7页
Journal of Southwest University(Natural Science Edition)
基金
国家自然科学基金项目(11401487)
关键词
无约束优化
谱共轭梯度法
全局收敛
Wolfe线搜素
unconstrained optimization
spectral conjugate gradient method
global convergence
Wolfe line search