摘要
本文给出确定共轭梯度方法中参数βk范围的两个条件-条件Ⅰ和条件Ⅱ,它们都确保方法的全局收敛.在条件Ⅰ和Gilbert&Nocedal(1992)引入的性质(*)下及在条件Ⅱ和Wolfe条件下,分别建立了共轭梯度算法的收敛性定理.
Two conditions, Ⅰ and Ⅱ, defining the range of parameter βk in a a conjugate gradient method, are given in the paper to ensure the global convergence. Convergence theorems are established under Condition Ⅰ and Property (*) introduced by Gilbert & Nocedal(1992), or condition Ⅱ and Wolfe conditions, respectively.
出处
《运筹学学报》
CSCD
1999年第2期71-81,共11页
Operations Research Transactions