本文介绍一种沿曲线方向线性搜索的算法，这种算法概括了梯度法、共轭梯度法以及它们的改进格式。
This paper introduces a curvilinear line searching technique, which possesses all the meris of the gradient method, the conjugate gradient method and some of their refined types.
文章给出了一种新的非精确线性搜索下的共轭梯度法，说明了在新线性搜索下每次迭代能够产生下降方向。
The method is proved to be globally convergent if the exact line search, the Grippo-Lucidi line search and the Wolfe-Powell line search is used.
应用推荐