广义Cury线搜索下一种新共轭梯度法及其全局收敛性

Global Convergence of a New Conjugate Gradient Algorithm with a Generalized Curry linesearch

本文对无约束最优化问题：ｍｉｎｆ（ｘ），ｘ∈Ｒｎ，提出一种新的共轭梯度法．该算法中参数βｋ采用一种新取法，并结合广义Ｃｕｒｙ线搜索及ｎ步重新开始策略．在关于目标函数较弱条件假设下，证明了所给算法的全局收敛性．

The paper proposes a new conjugate gradient algorithm for unconstrained optimization problem. This algorithm used the technique of the generalized Curry linessarch and the tactics of the n steps restarting, and a new formule for parameter β k is given, the global convergence of this algorithm is proved under weaker conditions.