非线性规划的一种摄动梯度投影

A Perturbation Gradient Projection Algorithm for Nonlinear Programming

本文应用对约束切子空间摄动的技巧,寻找一种新的极其简单的摄动参数,给出了一种在非K—T点处各种情形下目标函数的梯度都无须投影两次的算法,并证明了这种非线性规划新算法的收敛性。

In this paper,a new simple perturbation parameter is found with the technique of perturbing tangent subspace and a new algorithm requiring only one projection of the gradient of objective function at non K-T point in any case is presented. The proof of global convergence of this algrothm is given as well.