一种新共轭梯度法的全局收敛性

The global convergent property of a new conjugate gradient method

对求解无约束最优化问题的共轭梯度法进行了研究,提出了计算βk的一种新的公式,并对标准W o lfe搜索条件进行了推广,得到一种新的共轭梯度法。在一定条件下证明了该算法的全局收敛性,同时给出了一些数值例子,得到很好的数值结果。

This paper presents an exploration into the conjugate method for unconstrained optimization. A new conjugate method with global convergence for Bk has been proposed. This new method features a wider line search than the standard Wolfe search. In addition, numerical experiments have been made and good numerical results have been obtained.