非线性规划不用梯度作线搜索的算法及收敛性

Algorithms for Nonlinear Programming Without Using Gradient for Linesearch and It's Convergence

首先给出求解无约束规划中步长搜索依赖的几个准则和一个线搜索方法，它们仅利用目标函数值，不依赖于梯度。然后，给出点列整体收敛的充分条仲，并建立无约束规划的几个具有整体收敛性的算法。此外，还给出算法超线性收敛的条件。最后将方法应用于求解等式约束规划，产生其整体收敛的不用梯度的可行下降算法。

This paper first gives a linesearch mehtod and several criterias which thestepsize depends on in solving unconstrained minimization problem, them depend only onthe object function values, not the gradients. Then the sufficient condictions for globalconvergence are given, and several global convergent algorithms for unconstrainedminlmization are set up which this condictions, in addition to. the superlinear convergentconditions for algorithm are given.