本文研究无约束优化问题.利用前面多步迭代点的信息产生下降方向以及Armijo线性搜索产生步长,得到了一类新的多步下降算法,并且在较弱条件下证明了算法具有全局收敛性和线性收敛速率.初步的数值试验表明算法是有效的.
In this paper,the unconstrained optimization problem is investigated.By using the previous multi-step iterative information to generate a decent direction and applying Armijo linear search to define the step-size,we present a new multi-step decent algorithm and prove the global convergence and linear convergence rate of the algorithm under some mild conditions.Preliminary numerical results demonstrate the effectiveness of the algorithm.