Reorthogonalization for the Golub-Kahan-Lanczos bidiagonal reduction

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

The Golub-Kahan-Lanczos (GKL) bidiagonal reduction generates, by recurrence, the matrix factorization of X ∈ ℝm×n, m≥n, given by X = UBVT where U ∈ ℝm×n is left orthogonal, V ∈ ℝm×n is orthogonal, and B ∈ ℝm×n is bidiagonal. When the GKL recurrence is implemented in finite precision arithmetic, the columns of U and V tend to lose orthogonality, making a reorthogonalization strategy necessary to preserve convergence of the singular values. The use of an approach started by Simon and Zha (SIAM J Sci Stat Comput, 21:2257-2274, 2000) that reorthogonalizes only one of the two left orthogonal matrices U and V is shown to be very effective by the results presented here. Supposing that V is the matrix reorthogonalized, the reorthogonalized GKL algorithm proposed here is modeled as the Householder Q-R factorization of, where Vk = V(:,1:k). That model is used to show that if εM is the machine unit and, where tril(·) is the strictly lower triangular part of the contents, then: (1) the GKL recurrence produces Krylov spaces generated by a nearby matrix X + δX, {double pipe}δ X{double pipe}F = O (εM + n̄) {double pipe}X{double pipe}F; (2) singular values converge in the Lanczos process at the rate expected from the GKL algorithm in exact arithmetic on a nearby matrix; (3) a new proposed algorithm for recovering leading left singular vectors produces better bounds on loss of orthogonality and residual errors.

Original languageEnglish (US)
Pages (from-to)237-278
Number of pages42
JournalNumerische Mathematik
Volume124
Issue number2
DOIs
StatePublished - May 1 2013

Fingerprint

Lanczos
Recurrence
Lanczos Algorithm
Pipe
Singular Values
Orthogonality
Factorization
Factorization of Matrices
QR Factorization
Singular Vectors
Orthogonal matrix
Triangular
Strictly
Tend
Converge
Unit
Necessary

All Science Journal Classification (ASJC) codes

  • Computational Mathematics
  • Applied Mathematics

Cite this

@article{ec215a82736e4ca396ae4616d85ca8c0,
title = "Reorthogonalization for the Golub-Kahan-Lanczos bidiagonal reduction",
abstract = "The Golub-Kahan-Lanczos (GKL) bidiagonal reduction generates, by recurrence, the matrix factorization of X ∈ ℝm×n, m≥n, given by X = UBVT where U ∈ ℝm×n is left orthogonal, V ∈ ℝm×n is orthogonal, and B ∈ ℝm×n is bidiagonal. When the GKL recurrence is implemented in finite precision arithmetic, the columns of U and V tend to lose orthogonality, making a reorthogonalization strategy necessary to preserve convergence of the singular values. The use of an approach started by Simon and Zha (SIAM J Sci Stat Comput, 21:2257-2274, 2000) that reorthogonalizes only one of the two left orthogonal matrices U and V is shown to be very effective by the results presented here. Supposing that V is the matrix reorthogonalized, the reorthogonalized GKL algorithm proposed here is modeled as the Householder Q-R factorization of, where Vk = V(:,1:k). That model is used to show that if εM is the machine unit and, where tril(·) is the strictly lower triangular part of the contents, then: (1) the GKL recurrence produces Krylov spaces generated by a nearby matrix X + δX, {double pipe}δ X{double pipe}F = O (εM + n̄) {double pipe}X{double pipe}F; (2) singular values converge in the Lanczos process at the rate expected from the GKL algorithm in exact arithmetic on a nearby matrix; (3) a new proposed algorithm for recovering leading left singular vectors produces better bounds on loss of orthogonality and residual errors.",
author = "Barlow, {Jesse Louis}",
year = "2013",
month = "5",
day = "1",
doi = "10.1007/s00211-013-0518-8",
language = "English (US)",
volume = "124",
pages = "237--278",
journal = "Numerische Mathematik",
issn = "0029-599X",
publisher = "Springer New York",
number = "2",

}

Reorthogonalization for the Golub-Kahan-Lanczos bidiagonal reduction. / Barlow, Jesse Louis.

In: Numerische Mathematik, Vol. 124, No. 2, 01.05.2013, p. 237-278.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Reorthogonalization for the Golub-Kahan-Lanczos bidiagonal reduction

AU - Barlow, Jesse Louis

PY - 2013/5/1

Y1 - 2013/5/1

N2 - The Golub-Kahan-Lanczos (GKL) bidiagonal reduction generates, by recurrence, the matrix factorization of X ∈ ℝm×n, m≥n, given by X = UBVT where U ∈ ℝm×n is left orthogonal, V ∈ ℝm×n is orthogonal, and B ∈ ℝm×n is bidiagonal. When the GKL recurrence is implemented in finite precision arithmetic, the columns of U and V tend to lose orthogonality, making a reorthogonalization strategy necessary to preserve convergence of the singular values. The use of an approach started by Simon and Zha (SIAM J Sci Stat Comput, 21:2257-2274, 2000) that reorthogonalizes only one of the two left orthogonal matrices U and V is shown to be very effective by the results presented here. Supposing that V is the matrix reorthogonalized, the reorthogonalized GKL algorithm proposed here is modeled as the Householder Q-R factorization of, where Vk = V(:,1:k). That model is used to show that if εM is the machine unit and, where tril(·) is the strictly lower triangular part of the contents, then: (1) the GKL recurrence produces Krylov spaces generated by a nearby matrix X + δX, {double pipe}δ X{double pipe}F = O (εM + n̄) {double pipe}X{double pipe}F; (2) singular values converge in the Lanczos process at the rate expected from the GKL algorithm in exact arithmetic on a nearby matrix; (3) a new proposed algorithm for recovering leading left singular vectors produces better bounds on loss of orthogonality and residual errors.

AB - The Golub-Kahan-Lanczos (GKL) bidiagonal reduction generates, by recurrence, the matrix factorization of X ∈ ℝm×n, m≥n, given by X = UBVT where U ∈ ℝm×n is left orthogonal, V ∈ ℝm×n is orthogonal, and B ∈ ℝm×n is bidiagonal. When the GKL recurrence is implemented in finite precision arithmetic, the columns of U and V tend to lose orthogonality, making a reorthogonalization strategy necessary to preserve convergence of the singular values. The use of an approach started by Simon and Zha (SIAM J Sci Stat Comput, 21:2257-2274, 2000) that reorthogonalizes only one of the two left orthogonal matrices U and V is shown to be very effective by the results presented here. Supposing that V is the matrix reorthogonalized, the reorthogonalized GKL algorithm proposed here is modeled as the Householder Q-R factorization of, where Vk = V(:,1:k). That model is used to show that if εM is the machine unit and, where tril(·) is the strictly lower triangular part of the contents, then: (1) the GKL recurrence produces Krylov spaces generated by a nearby matrix X + δX, {double pipe}δ X{double pipe}F = O (εM + n̄) {double pipe}X{double pipe}F; (2) singular values converge in the Lanczos process at the rate expected from the GKL algorithm in exact arithmetic on a nearby matrix; (3) a new proposed algorithm for recovering leading left singular vectors produces better bounds on loss of orthogonality and residual errors.

UR - http://www.scopus.com/inward/record.url?scp=84877769119&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84877769119&partnerID=8YFLogxK

U2 - 10.1007/s00211-013-0518-8

DO - 10.1007/s00211-013-0518-8

M3 - Article

AN - SCOPUS:84877769119

VL - 124

SP - 237

EP - 278

JO - Numerische Mathematik

JF - Numerische Mathematik

SN - 0029-599X

IS - 2

ER -