一个新的共轭投影梯度算法及其超线性收敛性

A NEW CONJUGATE PROJECTION GRADIENT METHOD AND SUPERLINEAR CONVERGENCE

利用共轭投影梯度技巧,结合SQP算法的思想,建立了一个具有显示搜索方向的新算法.在适当的条件下,证明算法是全局收敛和强收敛的,且具有超线性收敛性.最后数值实验表明算法是有效的.

In this paper, a new method with explicit direction is presented by combining the concept of conjugate projection gradient with the idea of SQP method, and its global convergence and superlinear convergence are obtained under certain conditions. The numerical results show that the method in this paper is effective.