一种充分下降的共轭梯度法

A Sufficient Descent Conjugate Gradient Method

在Dai-Liao共轭梯度法的基础上,提出了一种修正的共轭梯度法,该算法在强Wolfe线性搜索和精确线性搜索下具有充分下降性.同时,在确定步长的过程中,如果出现某个步长很小,则该算法的搜索方向会自动的接近当前迭代点的负梯度方向.

Based on the Dai-Liao conjugate gradient method, a modified conjugate gradient method is presented.Under the strong Wolfe linear search or exact linear search, the direction generated by the proposed method is always descent direction. An attractive property is that when a small step length occurs, the search direction automatically approaches to the steepest descent direction.