求解大规模无约束优化问题的一种新的PRP三项共轭梯度法

A New Modified PRP Conjugate Gradient Method for Solving Large-scale Unconstrained Problems

提出了一种修正Polak—Ribiere—Polyak(PRP)三项共轭梯度算法,在Yuan-Wei-Lu不精确线搜索下,求解大规模无约束优化问题.在适当的条件下,新算法具有充分下降性和信赖域特征,对于非凸函数具有全局收敛性.初步的数值实验表明,新算法比相似算法更有效.

In this paper,we proposed a modified PRP conjugate gradient method for solving large-scale unconstrained problems,where this paper utilizes Yuan-Wei-Lu nonexact line search technique.With some suitable conditions,the new algorithm has the following properties:automatically with sufficient descent property,trust region feature,global convergence for nonconvex functions;and The ideal numerical performance compare with similar algorithms.