摘要
在修正的拟牛顿方程的基础上,给出了一种适用于求解大规模问题的有限内存对称秩一算法.该算法充分 利用了迭代过程所得到的函数值和相应的梯度值.同时,用有限内存技术改造一般对称秩一算法,给出了对称秩一 矩阵的有限内存矩阵表示,从而大大节省了计算机的内存和计算量,使算法更适用于大规模优化问题的求解.
Based on the modified quasi-Newton equation, a limited memory symmetric rank 1 (L-HSR1) algorithm is given.The presented method makes more available information on both the function and the gradient, in which, both of them are employed to increase the accuracy of Hessian approximations. Moreover, the limited memory technique is combined with the HSR1 quasi-Newton method. A matrix presentation of limited memory SR1(Symmetric Rank 1) update is derived so that EMS memory and computation are saved. As a result, the presented method is more fit for large scale optimization.
出处
《宁夏大学学报（自然科学版）》
CAS
2004年第3期219-222,共4页
Journal of Ningxia University(Natural Science Edition)
基金
国家自然科学基金资助项目(10231060)
