共轭梯度法的一个计算公式

共轭梯度法是最典型的共轭方向法。文中给出了另一种有效的共轭梯度法，它适用于求解目标函数为一般可微函数的无约束最优化问题，而且其中的线性搜索不必用精确线性搜索，只需满足Ｗｏｌｆｅ准则。文中对该方法的收敛性给予了分析，同时还讨论了具体实现方法，并用经典算例进行了数据试验。

Calculation Formula of Conjugate Gradient Method

The conjugate direction method is one between the gradient method and Newton methods. The method only needs the first order derivative of objective function ,and converges more quick than the gradient method. Conjugate gradient (CG) method is a most typical one in the Conjugate Direction methods. This paper gives another efficient algorithm of CG,in which the line search scheme may satisfy the Wolfe criterion. This algorithm can be applied to the general unconstrained optimization. The paper also analyses convergence and implementation of this algorithm. Finally a numerical experiment is presented.