摘要
在相关文献的基础上,提出无约束优化问题的修正WYL共轭梯度法,该方法不依赖于任何线搜索而满足充分下降性,证明了采用Arimijo型线搜索算法的全局收敛性,及在适当条件下算法具有R-线收敛速率,证明了采用重开始策略时算法具有n步二次收敛性。
Based on the relevant literatures, a new modified WYL conjugate gradient method for unconstrained optimization is proposed. The new method has sufficient descent property without any line search condition. Under Armijo-type line search, this paper shows that the new method possess- es global convergence and linear convergence rate. The new method is quadratic ally convergent with a restart strategy if the initial step length is appro- priate.
基金
广西教育厅科研项目(201012MS215)
关键词
无约束优化
共轭梯度法
全局收敛
重开始策略
n步二次收敛
unconstrained optimization
conjugate gradient method
global convergence
restart strategy
n-step quadratic convergence
