用网络实现矩阵奇异值分解

Network Method for Singular Value Decomposition

用网络求实对称矩阵的特征值及其相应的特征向量。从而实现矩阵的奇异值分 解。在只需求出几个较大特征值的情况下，这种方法比较简单并易于并行实现。文中还 提出逐步求矩阵的特征值和特征向量的剥去法。给出了有关证明和算例。

This paper finds the eigenvalues and eigenvectors of a real symmetric matrix with a network method for the purpose of calculating singular value decomposition. For cases with only a few larger eigenvalues to be determined, this method is simpler and can be completed easily with- parallel algorithm. The paper proposes also a peeling method used to find matrix eigenvalues and eigenvectors in successive steps. Related proofs and examples are given.