无约束最优化的一种新的拟Newton型法

A New Quasi-Newton Type Method for Unconstrained

本文讨论了一种新的拟Newton型法。本算法在迭代过程的每一步中只修正对角元,以产生新的校正矩阵,校正矩阵可保持对称性和稀疏性,并尽量满足拟Newton方程。在一定的条件下,本算法是局部超线性收敛的。本文中一些数值例子也说明该算法是可信赖的。

A new Quasi-Newton type me thod is discussed in this paper.This method only changes diagonal factor to construct new updating matrix in every iteration. Therefore, the updating matrix can keep the sparsity and the symmetry. Certainly, the Quasi-Newton Fundamental Equation is satisfied as far as possible, Under some conditions, this method is local and superlinear convergent. Some numerical experiments seem to confirm that the new algorithm is reliable.