摘要
通过对不同共轭梯度法收敛性分析的研究,提出了共轭梯度法全局收敛的一个充分条件,分析了该充分条件的合理性,并给出一种带参数的混合共轭梯度法,证明了该方法在强Wolfe线搜索下满足该充分条件.数值实验结果表明:该算法是有效的.
On the basis of research on convergence of different conjugate gradient methods, a sufficient condition of global convergence for one conjugate gradient method was proposed. The rationality of the sufficient condition was analyzed, and a new hybrid conjugate gradient method with two parameters was presented. It was proved that the method in the strong Wolfe line search meets this sufficient condition. Numerical results show that the method is efficient.
出处
《中北大学学报（自然科学版）》
CAS
北大核心
2010年第1期5-8,共4页
Journal of North University of China(Natural Science Edition)
关键词
无约束优化
共轭梯度法
线搜索
全局收敛性
充分条件
unconstrained optimization
conjugate gradient method
line search
global convergence
sufficient condition
