Abstract:SIAM Journal on Matrix Analysis and Applications, Volume 45, Issue 1, Page 24-58, March 2024. The cross-product matrix-based CJ-FEAST SVDsolver proposed previously by the authors is shown to compute the left singular vector possibly much less accurately than the right singular vector and may be numerically backward unstable when a desired singular value is small. In this paper, an alternative augmented matrix-based CJ-FEAST SVDsolver is proposed to compute the singular triplets of a large matrix [math] with the singular values in an interval [math] contained in the singular spectrum. The new CJ-FEAST SVDsolver is a subspace iteration applied to an approximate spectral projector of the augmented matrix [math] associated with the eigenvalues in [math], and it constructs approximate left and right singular subspaces independently, onto which [math] is projected to obtain the Ritz approximations to the desired singular triplets. Compact estimates are given for the accuracy of the approximate spectral projector constructed by the Chebyshev–Jackson series expansion in terms of series degree, and a number of convergence results are established. The new solver is proved to be always numerically backward stable. A convergence comparison of the cross-product-based and augmented matrix-based CJ-FEAST SVDsolvers is made, and a general-purpose choice strategy between the two solvers is proposed for the robustness and overall efficiency. Numerical experiments confirm all the results and meanwhile demonstrate that the proposed solver is more robust and substantially more efficient than the corresponding contour integral-based versions that exploit the trapezoidal rule and the Gauss–Legendre quadrature to construct an approximate spectral projector.

An algorithm for J-spectral factorization of certain matrix functions

A numerical algorithm for matrix spectral factorization on the real line

On multivariable matrix spectral factorization method

A New Method of Matrix Spectral Factorization

On the exact spectral factorization of rational matrix functions with applications to paraunitary filter banks

THE CALCULATION OF J_SPECTRUM FACTORIZATION FOR THE FULL RANK PARA-HERMITE POLYNOMIAL MATRIX

Spectral Factorization of Rank-Deficient Polynomial Matrix-Functions

A Matrix Polynomial Spectral Approach for General Joint Block Diagonalization.

Solvability of an Inverse Spectral Problem for Jacobi Matrices with Mixed Given Data

An Augmented Matrix-Based CJ-FEAST SVDsolver for Computing a Partial Singular Value Decomposition with the Singular Values in a Given Interval

Inverse eigenvalue problem for pseudo-symmetric Jacobi matrices with two spectra

A sparse-sparse iteration for computing a sparse incomplete factorization of the inverse of an SPD matrix

An Inverse Spectral Problem for Non-Self-Adjoint Jacobi Matrices

Enhanced shifted Jacobi operational matrices of integrals: spectral algorithm for solving some types of ordinary and fractional differential equations

A Class of Inexact Block Factorization Preconditioners for Indefinite Matrices with a Three-by-three Block Structure.

Spectral Decomposition of Real Symmetric Quadratic Λ-Matrices and Its Applications.

Iterative methods for solving factorized linear systems

A FEAST SVDsolver Based on Chebyshev–Jackson Series for Computing Partial Singular Triplets of Large Matrices

The Joint Bidiagonalization Method for Large GSVD Computations in Finite Precision.

Some Remarks on the Complex J-symmetric Eigenproblem

A Structure Preserving Matrix Factorization for Solving General Periodic Pentadiagonal Toeplitz Linear Systems.