| 摘要: |
| 基于信赖域技术和修正拟牛顿方程,结合Neng-Zhu Gu非单调策略,设计新的求解无约束最优化问题的非单调超记忆梯度算法,分析算法的收敛性和收敛速度。新算法每次迭代节约了矩阵的存储量和计算量,算法稳定, 适于求解大规模问题。数值试验结果表明新算法是有效的。 |
| 关键词: 超记忆梯度算法 非单调规则 收敛性 收敛速度 数值试验 |
| DOI：10.3969/j.issn.1673-5005.2013.02.032 |
| 分类号::O 221.2 |
| 基金项目:国家自然科学基金项目(61201455);中央高校基本科研业务费专项(10CX04044A;11 CX06087A) |
|
| A non-monotone super-memory gradient method based on trust region technique and modified quasi-Newton equation |
|
GONG En-long1, CHEN Shuang-shuang2, SUN Qing-ying2, CHEN Ying-mei2
|
|
(1.Qingdao Hotel Management College, Qingdao 266100, China;2.College of Science in China University of Petroleum, Qingdao 266580, China)
|
| Abstract: |
| Based on trust region technique and modified quasi-Newton equation, by combining with Neng-Zhu Gu non-monotone strategy, a new super-memory gradient method for unconstrained optimization problem was presented. The global and convergence properties of the new method were proved. It saves the storage and computation of some matrixes in its iteration, and is suitable for solving large scale optimization problems. The numerical results show that the new method is effective. |
| Key words: super-memory gradient method non-monotone step rule convergence convergence rate numerical experiment |
