基于非线性优化中的截断牛顿法提出了解大型稀疏对称矩阵特征值问题的一个子空间加速的截断牛顿法,证明了算法的收敛性并进行了数值试验,数值试验结果表明数值结果与理论分析相符,表明该算法是有效的。
Based on the classic truncated Newton algorithm for nonlinear optimization, a new subspace accelerated trun- cated Newton algorithm is presented for large-scale symmetric sparse extreme eigenvalue problems. The convergence properties are proved and numerical experiments are done. The numerical results match the convergence analysis, which shows that the new algorithm is effective.