Generalized epsilon-pseudospectra
- New York Univ., NY (United States). Courant Inst. of Mathematical Sciences
Normal matrices have complete sets of orthonormal eigenvectors, and therefore the spectral decomposition is useful in studying the properties of normal operators. In contrast, the eigenvectors of nonnormal matrices can be nearly linearly dependent, and the eigenvalue problem may by highly ill conditioned. Thus additional concepts and analysis techniques are useful in examining nonnormal operators. The concepts of [epsilon]-pseudospectra, introduced by L.N. Trefethen, has proven to be a powerful tool in the analysis of nonnormal operators. The author generalizes [epsilon]-pseudospectra and the associated computational algorithms to the generalized eigenvalue problem. Rank 1 perturbations are used to determine the [epsilon]-pseudospectra.
- DOE Contract Number:
- FG02-86ER53223
- OSTI ID:
- 6999912
- Journal Information:
- SIAM Journal on Numerical Analysis (Society for Industrial and Applied Mathematics); (United States), Journal Name: SIAM Journal on Numerical Analysis (Society for Industrial and Applied Mathematics); (United States) Vol. 31:4; ISSN 0036-1429; ISSN SJNAAM
- Country of Publication:
- United States
- Language:
- English
Similar Records
Iterative methods for solving Ax=b, GMRES/FOM versus QMR/BiCG