一类共轭梯度算法的全局收敛性(英文)

Golbal Convergence for A Class of Conjugate Gradient Methods

本文证明了一类共轭梯度算法的全局收敛性，其中参数βk满足|βk|≤β，并且αk满足放宽了的强Wolfe线搜索（max｛σ1，σ2｝≤1／2）．

This paper proves the convergence of a class of conjugae gradient methods, in which parameter βk satisfies |βk|≤β and the relaxed strong Wolfe line-search, with max{σ1, σ2} ≤ 1/2, is adopted.