摘要
在共轭梯度不能精确计算的情况下,研究了带误差项的FR共轭梯度法,因而更适用于许多实际问题.步长规则采用强Wolfe线搜索,在很一般的条件下证明了算法的全局收敛性,并给出了数值实验.结果表明,误差项的出现不影响算法的稳定性.
In the case where the conjugate gradient is computed inexactly, we con- sider FR conjugate gradient method with errors, which is more applicable in practice. For the step size rule, the strong Wolfe line search is used, and the global conver- gence of the method is proven under very general conditions. Finally, the numerical experiment results indicate that errors are independent of the stability of algorithm.
出处
《系统科学与数学》
CSCD
北大核心
2013年第10期1135-1143,共9页
Journal of Systems Science and Mathematical Sciences
基金
国家自然科学基金(71261010)资助课题
关键词
共轭梯度法
强Wolfe线搜索
全局收敛性
误差
Conjugate gradient method, strong Wolfe line search, global conver-gence, error.
