| 摘要: |
| 针对无约束优化问题，利用两项共轭梯度法(DL方法)去逼近改进的HS三项共轭梯度法，提出了改进的DL共轭梯度法即MDL共轭梯度法.该方法相对于DL方法具有一个更好的性质，即该共轭梯度法的搜索方向不依赖任何线搜索就可满足充分下降条件,理论上证明了该方法在Wolfe线搜索条件下对一般函数具有全局收敛性. |
| 关键词: 共轭梯度法 Wolfe线搜索 充分下降性 全局收敛性 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| A Class of DL Conjugate Gradient Method with Sufficient Descent |
|
XIE Li
|
| Abstract: |
| For unconstrained optimization problems, the improved HS three-term conjugate gradient method is approximated by the two-term conjugate gradient method DL method, and the improved DL conjugate gradient method MDL conjugate gradient method is proposed. Compared with DL conjugate gradient method, this method has a better property, that is, the search direction of the conjugate gradient method can satisfy sufficient descent condition without relying on any line search. It is theoretically proved that the method has global convergence for general functions under Wolfe line search condition. |
| Key words: conjugate gradient method Wolfe line search sufficient descent property global convergence |
