无约束优化的一类共轭梯度法(英)

A Class of Conjugate Gradient Methods for Unconstrained Optimization

本文给出了一类具有4个参数的共轭梯度法，并且分析了其中两个子类的方法．证明了在步长满足更一般的Wolfe条件时，这两个子类的方法是下降算法．同时还证明了这两个子类算法的全局收敛性．

In this paper, a class of conjugate gradient methods with 4 parameters in the choice ofthe scalar P. is presented and two subclasses of these methods are analysed. It is shown that these two subclasses of methods are descent methods when steplengths satisfy general Wolf e conditions. the global convergence of these two subclasses of the methods are also proved.