Bisection eigenvalue method for Hermitian matrices with quasiseparable representation and a related inverse problem
Eigenvalues, singular values, and eigenvectors (15A18) Inverse problems in linear algebra (15A29) Norms of matrices, numerical range, applications of functional analysis to matrix theory (15A60) Hermitian, skew-Hermitian, and related matrices (15B57) Numerical computation of eigenvalues and eigenvectors of matrices (65F15) Numerical computation of matrix norms, conditioning, scaling (65F35)
This paper builds off of the bisection method for computing eigenvalues of a matrix $A$. This is an iterative method that can find any eigenvalue of a real, symmetric, tridiagonal matrix based on the Sturm sequence property: for any given real number $\lambda$, the number of sign changes in the sequence of the characteristic polynomials $\gamma_0(\lambda)\equiv 1, \gamma_1(\lambda), \gamma_2(\lambda), \dots, \gamma_N(\lambda)$ of the principal leading submatrices of $A$ is the number of eigenvalues that are less than $\lambda$. The corresponding eigenvectors can also be computed. The authors study the bisection method for Hermitian matrices with quasiseparable representation. This extends their work [ETNA, Electron. Trans. Numer. Anal. 44, 342--366 (2015; Zbl 1332.65049)] for quasiseparable matrices of order one: here, the authors present an algorithm for using this method for Hermitian matrices whose quasiseparable generators of any order are known. This can be modified for a general matrix $A_0$ having a given quasiseparable representation, where the algorithm is applied to the Hermitian matrix $A=A_0^*A_0$. In order to check the accuracy and perform numerical tests of the developed eigenvalue algorithm, one requires a set of matrices with prescribed eigenvalues from which one can obtain the quasiseparable generators without building the entire matrix. The authors present an algorithm for solving this inverse problem. Various numerical experiments are performed with respect to errors as well as speed of the algorithm. For the entire collection see [Zbl 1411.47002].
- The bisection eigenvalue method for unitary Hessenberg matrices via their quasiseparable structure
- The fast bisection eigenvalue method for Hermitian order one quasiseparable matrices and computations of norms
- Solving the real eigenvalues of Hermitian quadratic eigenvalue problems via bisection
- The inverse eigenvalue problem for symmetric quasi anti-bidiagonal matrices
- The inverse eigenvalue problem for Hermitian anti-reflexive matrices and its approximation
- On solvability of inverse eigenvalue problems with Hermitian matrices
- scientific article; zbMATH DE number 25590
- scientific article; zbMATH DE number 1282036
- Publication:4935904
- The inverse generalized eigenvalue problem and the optimal approximation for Hermitian-reflexive matrices
- Direct and inverse eigenvalue problems for diagonal-plus-semiseparable matrices
- Divide and conquer method for eigenstructure of quasiseparable matrices using zeroes of rational matrix functions
- The bisection eigenvalue method for unitary Hessenberg matrices via their quasiseparable structure
- Improved bisection eigenvalue method for band symmetric Toeplitz matrices
- The fast bisection eigenvalue method for Hermitian order one quasiseparable matrices and computations of norms
This page was built for publication: Bisection eigenvalue method for Hermitian matrices with quasiseparable representation and a related inverse problem
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2010509)
