修正 LS 共轭梯度方法及其收敛性
On Modified LS Conjugate Gradient Method with Its Global Convergence
-
摘要: 提出了一种有效的修正LS共轭梯度方法。该方法在每一步迭代中均产生一个充分下降方向,且不依赖于任何线搜索。在强 Wolfe线搜索下,讨论了新方法对一般目标函数的全局收敛性。最后,与著名的 PRP方法、CG‐DESCENT方法比较,大量的数值试验表明,修正LS共轭梯度方法对给定的测试问题是有效的。Abstract: In this paper ,an efficient conjugate gradient method has been proposed based on the famous LS method .The search direction of the proposed method has the sufficient descent property ,which is inde‐pendent of any line search .Its global convergence for general functions under strong Wolfe line search has been discussed .Finally ,the numerical results show that the efficiency of the proposed method is encoura‐ging by comparing with PRP method and CG‐DESCENT method .
-
-
计量
- 文章访问数: 558
- HTML全文浏览数: 376
- PDF下载数: 2
- 施引文献: 0