English

Approximation Accuracy of the Krylov Subspaces for Linear Discrete Ill-Posed Problems

Numerical Analysis 2020-03-20 v3

Abstract

For the large-scale linear discrete ill-posed problem minAxb\min\|Ax-b\| or Ax=bAx=b with bb contaminated by Gaussian white noise, the Lanczos bidiagonalization based Krylov solver LSQR and its mathematically equivalent CGLS, the Conjugate Gradient (CG) method implicitly applied to ATAx=ATbA^TAx=A^Tb, are most commonly used, and CGME, the CG method applied to minAATyb\min\|AA^Ty-b\| or AATy=bAA^Ty=b with x=ATyx=A^Ty, and LSMR, which is equivalent to the minimal residual (MINRES) method applied to ATAx=ATbA^TAx=A^Tb, have also been choices. These methods exhibit typical semi-convergence feature, and the iteration number kk plays the role of the regularization parameter. However, there has been no definitive answer to the long-standing fundamental question: {\em Can LSQR and CGLS find 2-norm filtering best possible regularized solutions}? The same question is for CGME and LSMR too. At iteration kk, LSQR, CGME and LSMR compute {\em different} iterates from the {\em same} kk dimensional Krylov subspace. A first and fundamental step towards to answering the above question is to {\em accurately} estimate the accuracy of the underlying kk dimensional Krylov subspace approximating the kk dimensional dominant right singular subspace of AA. Assuming that the singular values of AA are simple, we present a general sinΘ\sin\Theta theorem for the 2-norm distances between these two subspaces and derive accurate estimates on them for severely, moderately and mildly ill-posed problems. We also establish some relationships between the smallest Ritz values and these distances. Numerical experiments justify the sharpness of our results.

Keywords

Cite

@article{arxiv.1805.10132,
  title  = {Approximation Accuracy of the Krylov Subspaces for Linear Discrete Ill-Posed Problems},
  author = {Zhongxiao Jia},
  journal= {arXiv preprint arXiv:1805.10132},
  year   = {2020}
}

Comments

33 pages, 11 figures. arXiv admin note: text overlap with arXiv:1701.05708, arXiv:1608.05907

R2 v1 2026-06-23T02:08:21.455Z