一种修正的Fletcher-Reeves共轭梯度法及其全局收敛性  

A Modified Fletcher-Reeves Conjugate Gradient Method and Its Global Convergence

作　　者：李敏[1] 

机构地区：[1]怀化学院数学与应用数学系,湖南怀化418008

出　　处：《西南师范大学学报（自然科学版）》2013年第3期14-19,共6页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：湖南省教育厅科研资助项目(12C0844)

摘　　要：提出了一种修正的Fletcher-Reeves共轭梯度法,该算法具有不依赖于线搜索方法的充分下降性.在一定条件下建立了基于Armijo或Wolfe线搜索利用该算法求解非凸优化问题的全局收敛性定理.相关的数值试验结果检验了该方法的有效性.In this paper,a modified Fletcher-Reeves conjugate gradient method has been proposed.The proposed method can generate sufficient directions for the objective function,and this property is independent of the line search used.Under mild conditions we show that the method convergent globally if the Armijo line search or Wolfe line search is used.The numerical results show that the proposed descent FR method is efficient.

关 键 词：Fletcher-Reeves方法 充分下降 ARMIJO搜索 WOLFE搜索 全局收敛 

分 类 号：O221.2[理学—运筹学与控制论]