摘要
强Wolfe条件不能保证标准CD共轭梯度法全局收敛.本文通过建立新的共轭参数,提出无约束优化问题的一个新谱共轭梯度法,该方法在精确线搜索下与标准CD共轭梯度法等价,在标准Wolfe线搜索下具有下降性和全局收敛性.初步的数值实验结果表明新方法是有效的,适合于求解非线性无约束优化问题.
Strong Wolfe line search conditions cannot guarantee the global convergence of standard CD conjugate gradi ent method. A new spectral conjugate gradient method for unconstrained optimization was proposed. This method is the same as the standard CD method when the line search is exact. Moreover, the corresponding algorithm was proved to be descent and globally convergent if the Wolfe line search is used. Preliminary numerical results show that the new method is efficent, suit able for solving nonlinear unconstrained optimization problems.
出处
《经济数学》
2013年第4期33-37,共5页
Journal of Quantitative Economics
基金
广西壮族自治区教育厅科研项目(201012MS215)
广西民族师范学院科研项目(2013RCGG002)
关键词
无约束优化
谱共轭梯度法
下降性
全局收敛
unconstrained optimization
spectral conjugate gradient method
descent property
global convergence