实规范矩阵的特征值问题

On the Eigenvalue Problem for Real Normal Matrices

中文摘要:

      

英文摘要:

      Using the important properties of normal matrices we generalize some algorithms which are efficient for solving symmetric eigenvalue problem.For small full normal matrices we generalize Jacobi algorithm to get Jacobi-like algorithm. For large sparse normal matrices A's we present two types of subspace iterative algorithm, one is derived from using symmetric subspace iterative algorithm to ATA. We also show that the convergence behavior of both types are same as that of symmetric subspace iterative algorithm. In addition, the generalization of summetric Lanczos algorithm is briefly discussed. All algorithms presented here for real matrices can be used by real arithmetic.