Spectral condition-number estimation of large sparse matrices.
From MaRDI portal
Abstract: We describe a randomized Krylov-subspace method for estimating the spectral condition number of a real matrix A or indicating that it is numerically rank deficient. The main difficulty in estimating the condition number is the estimation of the smallest singular value sigma_{min} of A. Our method estimates this value by solving a consistent linear least-squares problem with a known solution using a specific Krylov-subspace method called LSQR. In this method, the forward error tends to concentrate in the direction of a right singular vector corresponding to sigma_{min}. Extensive experiments show that the method is able to estimate well the condition number of a wide array of matrices. It can sometimes estimate the condition number when running a dense SVD would be impractical due to the computational cost or the memory requirements. The method uses very little memory (it inherits this property from LSQR) and it works equally well on square and rectangular matrices.
Recommendations
- A randomized algorithm for estimating the condition number of matrices
- scientific article; zbMATH DE number 434539
- Probabilistic bounds for the matrix condition number with extended Lanczos bidiagonalization
- scientific article; zbMATH DE number 94860
- Applications of statistical condition estimation to the solution of linear systems
Cited in
(11)- On choices of formulations of computing the generalized singular value decomposition of a large matrix pair
- Recovering a perturbation of a matrix polynomial from a perturbation of its first companion linearization
- A functional oriented truncation error adaptation method
- A new data mining approach to predicting matrix condition numbers
- scientific article; zbMATH DE number 434539 (Why is no real title available?)
- Probabilistic bounds for the matrix condition number with extended Lanczos bidiagonalization
- Applications of statistical condition estimation to the solution of linear systems
- scientific article; zbMATH DE number 94860 (Why is no real title available?)
- PRIMME\_SVDS: a high-performance preconditioned SVD solver for accurate large-scale computations
- A randomized algorithm for estimating the condition number of matrices
- Extreme singular values of inhomogeneous sparse random rectangular matrices
Describes a project that uses
Uses Software
This page was built for publication: Spectral condition-number estimation of large sparse matrices.
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5227181)
