强Wolfe线搜索下一种混合的PRP-WYL共轭梯度法

A Hybrid PRP-WYL Conjugate Gradient Method with the Strong Wolfe Line Search

【目的】为了结合Polak-Ribière-Polyak(PRP)共轭梯度法和Wei-Yao-Liu(WYL)共轭梯度法良好的理论和计算特性。【方法】通过分段函数,一种混合的PRP-WYL共轭梯度法被提出。【结果】在强Wolfe线搜索条件下,算法具有充分下降性和全局收敛性。【结论】初步数值结果表明,PRP-WYL算法比某些现有的包括PRP和WYL的共轭梯度算法更有效。

[Purposes]In order to combine nice theoretical and computational characteristics of the well-known Polak-Ribière-Polyak(PRP)conjugate gradient method and the Wei-Yao-Liu(WYL)conjugate gradient method.[Methods]By using a piecewise function,a hybrid PRP-WYL conjugate gradient method is proposed.[Findings]The sufficient descent property and global convergence of the proposed method was established under the strong Wolfe line search conditions.[Conclusions]Preliminary numerical results show that the PRP-WYL algorithm is more efficient and more robust than some of the existing conjugate gradient algorithms including the PRP and the WYL.