Low-Rank Matrix Approximations Do Not Need a Singular Value Gap
From MaRDI portal
Abstract: This is a systematic investigation into the sensitivity of low-rank approximations of real matrices. We show that the low-rank approximation errors, in the two-norm, Frobenius norm and more generally, any Schatten p-norm, are insensitive to additive rank-preserving perturbations in the projector basis; and to matrix perturbations that are additive or change the number of columns (including multiplicative perturbations). Thus, low-rank matrix approximations are always well-posed and do not require a singular value gap. In the presence of a singular value gap, connections are established between low-rank approximations and subspace angles.
Recommendations
- scientific article; zbMATH DE number 970009
- A study of the gap between the structured singular value and its convex upper bound for low-rank matrices
- Low-rank approximation of a matrix: novel insights, new progress, and extensions
- Lower bounds for the low-rank matrix approximation
- Singular Value Decompositions and Low Rank Approximations of Tensors
- Structured low rank approximation of non-negative matrices
- Singular value decomposition and the nearest matrix of a lower rank
- A minimum norm approach for low-rank approximations of a matrix
- Low-rank matrix approximation in the infinity norm
- Generalized low rank approximations of matrices
Cites work
- A note on element-wise matrix sparsification via a matrix-valued Bernstein inequality
- Accuracy and Stability of Numerical Algorithms
- An overview of relative \(\sin\Theta\) theorems for invariant subspaces of complex matrices
- Angles between subspaces and their tangents
- Error and Perturbation Bounds for Subspaces Associated with Certain Eigenvalue Problems
- Fast computation of low-rank matrix approximations
- Fast Monte Carlo Algorithms for Matrices I: Approximating Matrix Multiplication
- Fast Monte Carlo Algorithms for Matrices II: Computing a Low-Rank Approximation to a Matrix
- History and generality of the CS decomposition
- scientific article; zbMATH DE number 47363 (Why is no real title available?)
- scientific article; zbMATH DE number 194139 (Why is no real title available?)
- scientific article; zbMATH DE number 3799843 (Why is no real title available?)
- scientific article; zbMATH DE number 6125590 (Why is no real title available?)
- scientific article; zbMATH DE number 6159604 (Why is no real title available?)
- scientific article; zbMATH DE number 967931 (Why is no real title available?)
- New estimates for the recursive low-rank truncation of block-structured matrices
- Numerical methods for large eigenvalue problems
- On the numerical analysis of oblique projectors
- On the Rates of Convergence of the Lanczos and the Block-Lanczos Methods
- Perturbation bounds in connection with singular value decomposition
- Randomized approximation of the Gram matrix: exact computation and probabilistic bounds
- Sketching as a tool for numerical linear algebra
- Some new bounds on perturbation of subspaces
- Structural Convergence Results for Approximation of Dominant Subspaces from Block Krylov Spaces
- Subspace Iteration Randomization and Singular Value Problems
- The Rotation of Eigenvectors by a Perturbation. III
Cited in
(17)- Randomized block Krylov methods for approximating extreme eigenvalues
- Low phase-rank approximation
- Sensitivity of low-rank matrix recovery
- Perturbations of the \textsc{Tcur} decomposition for tensor valued data in the Tucker format
- A unified performance analysis of likelihood-informed subspace methods
- A study of the gap between the structured singular value and its convex upper bound for low-rank matrices
- Randomized subspace iteration: analysis of canonical angles and unitarily invariant norms
- A Theory of Quantum Subspace Diagonalization
- scientific article; zbMATH DE number 7307477 (Why is no real title available?)
- Perturbations of CUR Decompositions
- Admissible subspaces and the subspace iteration method
- A flexible PageRank-based graph embedding framework closely related to spectral eigenvector embeddings
- On the randomized SVD in infinite dimensions
- A novel adaptive low-rank matrix approximation method for image compression and reconstruction
- Accuracy and stability of CUR decompositions with oversampling
- Dominant subspace and low-rank approximations from block Krylov subspaces without a prescribed gap
- Computing Strong Rank-Revealing Factorizations for Matrices with Orthonormal Rows
This page was built for publication: Low-Rank Matrix Approximations Do Not Need a Singular Value Gap
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3119542)
