Some results on the regularization of LSQR for large-scale discrete ill-posed problems

被引:0
作者
Yi Huang
ZhongXiao Jia
机构
[1] Tsinghua University,Department of Mathematical Sciences
来源
Science China Mathematics | 2017年 / 60卷
关键词
ill-posed problem; regularization; Lanczos bidiagonalization; LSQR; CGLS; hybrid; 65F22; 65J20; 15A18;
D O I
暂无
中图分类号
学科分类号
摘要
LSQR, a Lanczos bidiagonalization based Krylov subspace iterative method, and its mathematically equivalent conjugate gradient for least squares problems (CGLS) applied to normal equations system, are commonly used for large-scale discrete ill-posed problems. It is well known that LSQR and CGLS have regularizing effects, where the number of iterations plays the role of the regularization parameter. However, it has long been unknown whether the regularizing effects are good enough to find best possible regularized solutions. Here a best possible regularized solution means that it is at least as accurate as the best regularized solution obtained by the truncated singular value decomposition (TSVD) method. We establish bounds for the distance between the k-dimensional Krylov subspace and the k-dimensional dominant right singular space. They show that the Krylov subspace captures the dominant right singular space better for severely and moderately ill-posed problems than for mildly ill-posed problems. Our general conclusions are that LSQR has better regularizing effects for the first two kinds of problems than for the third kind, and a hybrid LSQR with additional regularization is generally needed for mildly ill-posed problems. Exploiting the established bounds, we derive an estimate for the accuracy of the rank k approximation generated by Lanczos bidiagonalization. Numerical experiments illustrate that the regularizing effects of LSQR are good enough to compute best possible regularized solutions for severely and moderately ill-posed problems, stronger than our theory predicts, but they are not for mildly ill-posed problems and additional regularization is needed.
引用
收藏
页码:701 / 718
页数:17
相关论文
共 50 条
[21]   Simple stopping criteria for the LSQR method applied to discrete ill-posed problems [J].
Lothar Reichel ;
Hassane Sadok ;
Wei-Hong Zhang .
Numerical Algorithms, 2020, 84 :1381-1395
[22]   Fractional regularization matrices for linear discrete ill-posed problems [J].
Hochstenbach, Michiel E. ;
Noschese, Silvia ;
Reichel, Lothar .
JOURNAL OF ENGINEERING MATHEMATICS, 2015, 93 (01) :113-129
[23]   Minimization of functionals on the solution of a large-scale discrete ill-posed problem [J].
David R. Martin ;
Lothar Reichel .
BIT Numerical Mathematics, 2013, 53 :153-173
[24]   On Lanczos based methods for the regularization of discrete ill-posed problems [J].
Hanke, M .
BIT, 2001, 41 (05) :1008-1019A
[25]   On Lanczos Based Methods for the Regularization of Discrete Ill-Posed Problems [J].
Martin Hanke .
BIT Numerical Mathematics, 2001, 41 :1008-1018
[26]   Fractional regularization matrices for linear discrete ill-posed problems [J].
Michiel E. Hochstenbach ;
Silvia Noschese ;
Lothar Reichel .
Journal of Engineering Mathematics, 2015, 93 :113-129
[27]   The Regularized Global GMERR Method for Solving Large-Scale Linear Discrete Ill-Posed Problems [J].
Zhang, Hui ;
Dai, Hua .
EAST ASIAN JOURNAL ON APPLIED MATHEMATICS, 2024, 14 (04) :874-894
[28]   An l2-lq Regularization Method for Large Discrete Ill-Posed Problems [J].
Buccini, Alessandro ;
Reichel, Lothar .
JOURNAL OF SCIENTIFIC COMPUTING, 2019, 78 (03) :1526-1549
[29]   Regularization of exponentially ill-posed problems [J].
Hohage, T .
NUMERICAL FUNCTIONAL ANALYSIS AND OPTIMIZATION, 2000, 21 (3-4) :439-464
[30]   The Regularized Block GMERR Method and Its Simpler Version for Solving Large-Scale Linear Discrete Ill-Posed Problems [J].
Zhang, Hui ;
Dai, Hua .
COMMUNICATIONS ON APPLIED MATHEMATICS AND COMPUTATION, 2025, 7 (05) :1826-1847