无约束优化的下降谱共轭梯度算法  

A Descent Spectral Conjugate Gradient Algorithm for Unconstrained Optimization

作　　者：林穗华 LIN Sui-hua(College of Mathematics and Computer Sciences ,Guangxi Normal University .for Nationalities, Chongzuo 532200, Guangxi, Chin)

机构地区：[1]广西民族师范学院数学与计算机科学学院,广西崇左532200

出　　处：《内蒙古师范大学学报（自然科学汉文版）》2018年第1期1-6,共6页Journal of Inner Mongolia Normal University(Natural Science Edition)

基　　金：广西高校科研资助项目(ZD2014143);广西重点培育学科(应用数学)建设项目(桂教科研[2013]16);广西民族师范学院科研资助项目(2013RCGG002)

摘　　要：为求解无约束优化问题,设计含调比因子的共轭参数和谱参数,使相应的谱共轭梯度法具有下降性,在步长由Wolfe线搜索产生的情况下,新算法满足全局收敛性.初步的数值实验表明新算法是有效的.For solving unconstrained optimization problems,this paper designed a new conjugate parameter formula and appropriate spectral parameter formula with adjustable ratio factor,so that the spectral conjugate gradient method automatic satisfied the descent property.Under suitable conditions,the new algorithm with Wolfe line search was globally convergent.The numerical results showed that the new algorithm was effective.

关 键 词：无约束优化 谱共轭梯度法 WOLFE线搜索 下降性 全局收敛 

分 类 号：O221.2[理学—运筹学与控制论]