Structured eigenvalue condition numbers for parameterized quasiseparable matrices (Q728451): Difference between revisions
From MaRDI portal
Latest revision as of 03:33, 13 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Structured eigenvalue condition numbers for parameterized quasiseparable matrices |
scientific article |
Statements
Structured eigenvalue condition numbers for parameterized quasiseparable matrices (English)
0 references
20 December 2016
0 references
This paper deals with sensitivity analysis of eigenvalues for a class of low-rank structured matrices known as \(\{1;1\}\)-quasiseparable matrices. The structured condition numbers are defined and analysed with respect to relative perturbations of the parameters. The results show that some eigenvalues may be extremely ill-conditioned under general componentwise relative unstructured perturbations of the matrix entries, but very well-conditioned under perturbations in the parameters. Two remarkable unexpected properties are proved for the structured eigenvalue condition numbers with respect to the (infinitely many) quasiseparable representations. First, that these condition numbers are independent of the particular representation and, second, that they can be expressed just in terms of the matrix entries. The approach presented in this paper may be extended to other classes of low-rank structured matrices.
0 references
parameterized quasiseparable matrix
0 references
structued eigenvalue condition number
0 references
relative perturbation of parameters
0 references
fast algorithm
0 references
low-rank matrices
0 references
0 references
0 references
0 references
0 references
0 references