A graph-theoretic model of symmetric givens operations and its implications
暂无分享,去创建一个
Abstract Symmetric Givens operations ( A ′ ← GAG T , G a Givens rotation matrix) are a basic tool in many matrix computations, especially for eigenvalue-eigenvector computations. A graph-theoretic model of these operations is given for symmetric matrices, analogous to the graph-theoretic model of Cholesky factorization. Using this model, it is shown that unless there is “accidental cancellation,” it is impossible to reduce a range of different matrix classes to tridiagonal form in o ( n 2 ) Givens operations; these classes include arrowhead matrices, pentadiagonal matrices, and cyclic tridiagonal matrices.
[1] W. Gragg,et al. On computing accurate singular values and eigenvalues of matrices with acyclic graphs , 1992 .
[2] Gene H. Golub,et al. Matrix computations , 1983 .
[3] J. Pasciak,et al. Computer solution of large sparse positive definite systems , 1982 .
[4] G. Stewart,et al. Computing the eigenvalues and eigenvectors of symmetric arrowhead matrices , 1990 .
[5] Gene H. Golub,et al. QR-Like Algorithms for Symmetric Arrow Matrices , 1992, SIAM J. Matrix Anal. Appl..