| 引用本文： |
-
林穗华,黄海.一种新的杂交共轭梯度算法[J].广西科学,2007,14(4):378-380,385. [点击复制]
- LIN Sui-hua,HUANG Hai.A New Hybrid Conjugate Gradient Method[J].Guangxi Sciences,2007,14(4):378-380,385. [点击复制]
|
|
| 摘要: |
| 结合HS、DY和WYL方法提出求解无约束优化问题的共轭梯度公式中βk参数的一种新的计算公式:βk=(max{0,min{‖gk‖2,gkTyk-1,gkTŷk-1}})/(dk-1Tyk-1),并给出新的杂交共轭梯度算法;证明新算法在弱Wolf-Powell线搜索条件下具有全局收敛性,并用数值试验表明新算法具有较好的数值结果. |
| 关键词: 无约束优化 共轭梯度法 线搜索 全局收敛性 |
| DOI： |
| 投稿时间：2006-12-12修订日期：2007-03-26 |
| 基金项目:南宁师范高等专科学校科研项目(2007012)资助 |
|
| A New Hybrid Conjugate Gradient Method |
|
LIN Sui-hua1,2, HUANG Hai1
|
| (1.Department of Mathematics and Computer Science, Nanning Teachers College, Longzhou, Guangxi, 532400, China;2.Department of Statistics East China Normal University, Shanghai, 200062, China) |
| Abstract: |
| Combined with HS, DY and WYL methods, a new hybrid conjugate gradient method for unconstrained optimization is proposed.Its updated formula is βk= (max{0, min{‖gk‖2, gkTyk-1, gkTŷk-1}})/ (dk-1Tyk-1).We prove that, the corresponding method can ensure the global convergence under weak Wolf-Powell line search.Preliminary numerical results show that the proposed method is very efficient. |
| Key words: unconstrained optimination conjugate gradient method line search global convergence |
