methods for the eigenvalue problem with large sparse matrices

The eigenvalue problem Ax = XBx, where A and B are large and sparse symmetric matrices, is considered. An iterative algorithm for computing the smallest eigenvalue and its corresponding eigenvector, based on the successive overrelaxation splitting of the matrices, is developed, and its global convergence is proved. An ex- pression for the optimal overrelaxation factor is found in the case where A and B are two-cyclic (property A). Further, it is shown that this SOR algorithm is the first order approximation to the coordinate relaxation algorithm, which implies that the same overrelaxation can be applied to this latter algorithm. Several numerical tests are reported. It is found that the SOR method is more effective than coordinate relaxation. If the separation of the eigenvalues is not too bad, the SOR algorithm has a fast rate of convergence, while, for problems with more severe clustering, the c-g or Lanczos algorithms should be preferred.

[1]  C. Lanczos.  An iteration method for the solution of the eigenvalue problem of linear differential and integral operators , 1950 .

[2]  M. Hestenes,et al.  A method of gradients for the calculation of the characteristic roots and vectors of a real symmetric matrix , 1951 .

[3]  E. Stiefel,et al.  Refined Iterative Methods for Computation of the Solution and the Eigenvalues of Self-Adjoint Boundary Value Problems , 1959 .

[4]  W. Wasow,et al.  Finite-Difference Methods for Partial Differential Equations , 1961 .

[5]  R. Fletcher,et al.  New iterative methods for solution of the eigenproblem , 1966 .

[6]  W. Kahan.  Relaxation methods for an eigenproblem , 1966 .

[7]  G. Buffoni.  Evaluation of eigensolutions of discrete space diffusion equation , 1967 .

[8]  I. Fried.  Gradient methods for finite-element eigenproblems. , 1969 .

[9]  John K. Reid,et al.  The Solution of Large Sparse Unsymmetric Systems of Linear Equations , 1971, IFIP Congress.

[10]  Louis A. Hageman,et al.  Iterative Solution of Large Linear Systems. , 1971 .

[11]  Fred G. Gustavson,et al.  Some Basic Techniques for Solving Sparse Systems of Linear Equations , 1972 .

[12]  G. Stewart.  On the Sensitivity of the Eigenvalue Problem $Ax = \lambda Bx$ , 1972 .

[13]  C. Paige.  Computational variants of the Lanczos method for the eigenproblem , 1972 .

[14]  O. Axelsson.  A generalized SSOR method , 1972 .

[15]  Olof B. Widlund,et al.  On the use of Fast Methods for Separable Finite Difference Equations for the Solution of General Elliptic Problems , 1972 .

[16]  G. Golub,et al.  Use of fast direct methods for the efficient numerical solution of nonseparable elliptic equations. , 1972 .

[17]  C. Bender,et al.  The iterative calculation of several of the lowest or highest eigenvalues and corresponding eigenvectors of very large symmetric matrices , 1973 .

[18]  Paul E. Saylor,et al.  A Symmetric Factorization Procedure for the Solution of Elliptic Boundary Value Problems , 1973 .

[19]  H. Schwarz.  The eigenvalue problem (A − λB)x = 0 for symmetric matrices of high order , 1974 .

[20]  R. F. Hoskins,et al.  Computational Methods in Linear Algebra , 1976, The Mathematical Gazette.