pith. sign in

arxiv: 1907.08811 · v1 · pith:ZZCY3FKDnew · submitted 2019-07-20 · 🧮 math.NA · cs.NA

On the Golub--Kahan bidiagonalization for ill-posed tensor equations with applications to color image restoration

Fatemeh P. A. Beik , Khalide Jbilou , Mehdi Najafi-Kalyani , Lothar Reichel This is my paper

Pith reviewed 2026-05-24 18:50 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords tensor equationsGolub-Kahan bidiagonalizationTikhonov regularizationill-posed problemscolor image restorationStein tensor equationhigh-dimensional PDEs
0
0 comments X

The pith

Tensor Golub-Kahan bidiagonalization with Tikhonov regularization solves ill-posed tensor equations.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that the tensor Golub-Kahan bidiagonalization algorithm combined with Tikhonov regularization can solve ill-posed tensor linear equations. These equations arise from finite difference or collocation discretizations of high-dimensional problems. Theoretical results address the conditioning of Stein tensor equations and the applicability of the TGKB process to general cases. Numerical tests on classical problems and color image restoration illustrate the method's feasibility.

Core claim

The TGKB algorithm in conjunction with Tikhonov regularization can be exploited to solve ill-posed tensor linear equations. Theoretical results are presented to discuss the conditioning of the Stein tensor equation and to reveal how the TGKB process can be exploited for general tensor equations. Feasibility is shown through classical test problems and applications to color image restoration.

What carries the argument

The Tensor Golub--Kahan bidiagonalization (TGKB) algorithm, which produces a bidiagonal reduction suitable for applying Tikhonov regularization to tensor equations.

Load-bearing premise

The tensor linear equations obtained from finite difference or collocation discretizations of high-dimensional problems are sufficiently well-conditioned or structured for the TGKB process to produce useful regularized solutions without major instability or loss of information.

What would settle it

Running the TGKB-Tikhonov method on a standard color image restoration test case and finding that the restored image has higher error than a simple vectorized Tikhonov approach would challenge the claim of feasibility.

read the original abstract

This paper is concerned with solving ill-posed tensor linear equations. These kinds of equations may appear from finite difference discretization of high-dimensional convection-diffusion problems or when partial differential equations in many dimensions are discretized by collocation spectral methods. Here, we propose the Tensor Golub--Kahan bidiagonalization (TGKB) algorithm in conjunction with the well known Tikhonov regularization method to solve the mentioned problems. Theoretical results are presented to discuss on conditioning of the Stein tensor equation and to reveal that how the TGKB process can be exploited for general tensor equations. In the last section, some classical test problems are examined to numerically illustrate the feasibility of proposed algorithms and also applications for color image restoration are considered.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper proposes the Tensor Golub-Kahan bidiagonalization (TGKB) algorithm combined with Tikhonov regularization to solve ill-posed tensor linear equations arising from finite-difference or collocation discretizations of high-dimensional PDEs. Theoretical results address the conditioning of the Stein tensor equation and show how the TGKB process extends to general tensor equations. Numerical experiments on classical test problems and applications to color image restoration are used to demonstrate feasibility.

Significance. If the theoretical conditioning results and the numerical stability of TGKB hold, the work provides a structured extension of the classical Golub-Kahan process to the tensor setting, offering a practical regularization framework for high-dimensional ill-posed problems. The explicit treatment of the Stein equation and the image-restoration examples add concrete value; reproducible code or parameter-free derivations would further strengthen the contribution.

major comments (2)
  1. [§3] §3 (TGKB derivation): the reduction of the tensor equation to a bidiagonal form via the TGKB process is stated to preserve the essential singular-value information, but the multi-linear action of the tensor operator on the generated Krylov subspaces is not shown to be free of additional cross terms; a concrete bound relating the tensor residual to the bidiagonal residual is needed to support the claim that Tikhonov regularization on the reduced problem yields a reliable approximate solution.
  2. [§4] §4 (conditioning of Stein equation): the assertion that the Stein tensor equation is 'well-conditioned under mild assumptions' is used to justify applicability of TGKB, yet the proof sketch relies on a spectral-radius condition whose verification for the finite-difference discretizations in the numerical examples is not supplied; without this verification the central feasibility claim for the target class of problems remains open.
minor comments (3)
  1. Notation for the tensor-vector products and the unfolding operators is introduced without a consolidated table; a short notation summary would improve readability.
  2. Figure captions for the color-image examples do not state the regularization parameter selection method (e.g., discrepancy principle or L-curve); this detail is needed to reproduce the reported PSNR values.
  3. The abstract claims 'theoretical results are presented to discuss on conditioning,' but the introduction does not cite the precise theorem numbers that establish the conditioning bounds; cross-references should be added.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive suggestions. We respond to each major comment below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [§3] §3 (TGKB derivation): the reduction of the tensor equation to a bidiagonal form via the TGKB process is stated to preserve the essential singular-value information, but the multi-linear action of the tensor operator on the generated Krylov subspaces is not shown to be free of additional cross terms; a concrete bound relating the tensor residual to the bidiagonal residual is needed to support the claim that Tikhonov regularization on the reduced problem yields a reliable approximate solution.

    Authors: We agree that an explicit residual bound is required to rigorously justify the approach. The current manuscript states preservation of singular-value information but does not derive a concrete inequality controlling cross terms arising from the multilinear operator. In the revision we will add a theorem in §3 establishing ||r||_F ≤ C ||r_b|| where r is the tensor residual, r_b the bidiagonal residual, and C depends only on the norms of the orthogonal bases generated by TGKB. This will directly support the reliability of Tikhonov regularization on the reduced problem. revision: yes

  2. Referee: [§4] §4 (conditioning of Stein equation): the assertion that the Stein tensor equation is 'well-conditioned under mild assumptions' is used to justify applicability of TGKB, yet the proof sketch relies on a spectral-radius condition whose verification for the finite-difference discretizations in the numerical examples is not supplied; without this verification the central feasibility claim for the target class of problems remains open.

    Authors: The referee correctly notes that the spectral-radius condition is not numerically verified for the discretizations used. The theoretical discussion in §4 assumes the condition holds under mild assumptions, but the manuscript does not report the radii for the finite-difference operators appearing in the examples. In the revised version we will compute and tabulate these spectral radii for each test problem in §5, confirming that the radius is strictly less than one and thereby closing the feasibility gap. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper proposes the TGKB algorithm with Tikhonov regularization for ill-posed tensor equations arising from discretizations. It supplies its own theoretical results on Stein tensor equation conditioning and TGKB applicability to general tensor equations, plus numerical tests on classical problems and color image restoration. No load-bearing derivation step reduces by construction to a fitted input, self-definition, or self-citation chain; the claims rest on independent theory and external benchmarks within the paper's scope.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the method relies on standard numerical linear algebra assumptions not detailed here.

pith-pipeline@v0.9.0 · 5673 in / 1034 out tokens · 17849 ms · 2026-05-24T18:50:28.676612+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

  1. [1]

    A projection method to solve li near systems in tensor format

    Ballani J and Grasedyck L. A projection method to solve li near systems in tensor format. Numerical Linear Algebra with Applications. 2013; 20(1): 27–43

    work page 2013

  2. [2]

    MA TLAB Tensor Toolbox V ersion 2.5

    Bader BW and Kolda TG. MA TLAB Tensor Toolbox V ersion 2.5. http://www.sandia.gov/~tgkolda/TensorToolbox

    work page

  3. [3]

    On the Krylov subsp ace methods based on tensor format for positive definite Sylvester tensor equations

    Beik FPA, Movahed FS and Ahmadi-Asl S. On the Krylov subsp ace methods based on tensor format for positive definite Sylvester tensor equations. Numerical Linear Algebra with Applications. 2016; 23(3): 444–466

    work page 2016

  4. [4]

    El, Jbilou K and Reichel L

    Bentbib AH, Guide M. El, Jbilou K and Reichel L. A global La nczos method for image restoration, Journal of Computational and Applied Mathematics. 2016; 300 233–244

    work page 2016

  5. [5]

    El, Jbilou K and Reichel L

    Bentbib AH, Guide M. El, Jbilou K and Reichel L. Global Gol ubKahan bidiagonalization applied to large discrete ill-posed problems. Journal of Computational and Applied Mathematics. 2017; 322 46–56

    work page 2017

  6. [6]

    A generaliz ed global Arnoldi method for ill-posed matrix equations

    Bouhamidi A, Jbilou K, Reichel L, and Sadok H. A generaliz ed global Arnoldi method for ill-posed matrix equations. Journal of Computational and Applied Mathematics. 2012; 236 2078–2089

    work page 2012

  7. [7]

    Tikhonov–type iterative regularization met hods for ill-posed inverse problems:theoretical aspects and applications

    Buccini A. Tikhonov–type iterative regularization met hods for ill-posed inverse problems:theoretical aspects and applications. PhD thesis. University of Insubria. http://insubriaspace.cineca.it/bitstream/10277/703/1/Phd_Thesis_Buccinialessandro_completa.pdf

    work page

  8. [8]

    A projection method and Kronecker produc t preconditioner for solving Sylvester tensor equations

    Chen Z and Lu LZ. A projection method and Kronecker produc t preconditioner for solving Sylvester tensor equations. Science China Mathematics. 2012; 55(6): 1281–1292

    work page 2012

  9. [9]

    Nonnegative matrix and tensor factorizations: applicatio ns to exploratory multi-way data analysis and blind source separ ation

    Cichocki A, Zdunek R, Phan AH, Amari SI. Nonnegative matrix and tensor factorizations: applicatio ns to exploratory multi-way data analysis and blind source separ ation. John Wiley & Sons, 2009

    work page 2009

  10. [10]

    The Johns Hopkins University Press, Batimore

    Golub GH and V an V oan CF, Matrix Computations. The Johns Hopkins University Press, Batimore. 1996

    work page 1996

  11. [11]

    Matrix Analysis

    Horn RA and Johnson CR. Matrix Analysis. Cambridge University Press, Cambridge, UK, 1985

    work page 1985

  12. [12]

    On the choice of subspace fo r large-scale Tikhonov regularization problems in general form

    Huang G, Reichel L, and Yin F. On the choice of subspace fo r large-scale Tikhonov regularization problems in general form. Numerical Algorithms. in press. DOI:10.1007/s11075-018-0534-y

    work page doi:10.1007/s11075-018-0534-y

  13. [13]

    Krylov subspace methods to solve a class of tensor equations via the Einstein product Numerical Linear Algebra with Applications

    Huang B, Xie Y , and Ma C. Krylov subspace methods to solve a class of tensor equations via the Einstein product Numerical Linear Algebra with Applications. 2019; in press. DOI:10.1002/nla.2254

    work page doi:10.1002/nla.2254 2019

  14. [14]

    Tensor decompositions and applic ations

    Kolda TG and Bader BW. Tensor decompositions and applic ations. SIAM Review. 2009; 51(3): 455–500

    work page 2009

  15. [15]

    Low-rank tensor Krylov subspac e methods for parametrized linear systems

    Kressner D and Tobler C. Low-rank tensor Krylov subspac e methods for parametrized linear systems. SIAM Journal on Matrix Analysis and Applications. 2011; 32(1):1288–1316

    work page 2011

  16. [16]

    Sensitivity analysis of the Lyapuno v tensor equation

    Liang L and Zheng B. Sensitivity analysis of the Lyapuno v tensor equation. Linear Multilinear Algebra. 2019; 67(3):555–572

    work page 2019

  17. [17]

    Solving fully three-dim ensional microscale dual phase lag problem using mixed-collocation finite difference discretization

    Malek A, Bojdi ZK, Golbarg PNN. Solving fully three-dim ensional microscale dual phase lag problem using mixed-collocation finite difference discretization. Journal of Heat Transfer .2012; 134(9), 094504

    work page 2012

  18. [18]

    Mixed collocation–finite differe nce method for 3D microscopic heat transport prob- lems

    Malek A, Masuleh SHM. Mixed collocation–finite differe nce method for 3D microscopic heat transport prob- lems. Journal of Computational and Applied Mathematics. 2008; 217(1):137–147

    work page 2008

  19. [19]

    Viscoelastic flow in an undula ting tube using spectral methods

    Masuleh SHM, Phillips TN. Viscoelastic flow in an undula ting tube using spectral methods. Computers & Fluids. 2004; 33(8):1075–1095. 18 F. P. A. B EIK , K. J BILOU , M. N AJAFI -K ALYANI AND L. R EICHEL

    work page 2004

  20. [20]

    Chebyshev collocation spectral method for three-dimensional transient coupled radiative–conductive Heat transfer

    Sun YS, Jing M and Li BW. Chebyshev collocation spectral method for three-dimensional transient coupled radiative–conductive Heat transfer. Journal of Heat Transfer .2012; 134: 092701–092707

    work page 2012

  21. [21]

    On global iterat ive schemes based on Hessenberg process for (ill-posed) Sylvester tensor equations, Journal of Computational and Applied Mathematics

    Najafi-Kalyani M, Beik FPA and Jbilou K. On global iterat ive schemes based on Hessenberg process for (ill-posed) Sylvester tensor equations, Journal of Computational and Applied Mathematics. DOI: 10.1016/j.cam.2019.03.045

    work page doi:10.1016/j.cam.2019.03.045 2019

  22. [22]

    Extending BiCG and BiCR methods to sol ve the Stein tensor equation

    Xu X and Wang Q-.W. Extending BiCG and BiCR methods to sol ve the Stein tensor equation. Computers & Mathematics with Applications. 2019; 77(12): 3117–3127

    work page 2019

  23. [23]

    Nested splitting conjugate gra dient method for matrix equation AX B = C and preconditioning, Computers & Mathematics with Applications

    Zak MK and Toutounian F. Nested splitting conjugate gra dient method for matrix equation AX B = C and preconditioning, Computers & Mathematics with Applications. 2013; 66 269–278

    work page 2013

  24. [24]

    Regularization tools: A Matlab package for a nalysis and solution of discrete ill-posed prob- lems.Numerical Algorithms

    Hansen PC. Regularization tools: A Matlab package for a nalysis and solution of discrete ill-posed prob- lems.Numerical Algorithms. 1994; 6, 1–35. Software is available in http://www.netlib.org

    work page 1994