Abstract
Randomized Kaczmarz-type methods are appealing for large-scale linear systems arising from big data problems. One of the keys of randomized Kaczmarz-type methods is how to effectively select working rows from the coefficient matrix. To the best of our knowledge, most of the randomized Kaczmarz-type methods need to compute probabilities for choosing working rows. However, when the amount of data is huge, the computation of probabilities will be inaccurate due to many factors such as rounding errors and data distribution. Moreover, in some popular randomized Kaczmarz methods, we have to scan all the rows of the data matrix in advance, or to compute residual of the linear system in each step. Hence, we have to access all the rows of the data matrix, which are unfavorable for big data problems. To overcome these difficulties, we first introduce a semi-randomized Kaczmarz method in which there is no need to compute probabilities explicitly. However, we still have to access all the rows of the matrix for the computation of residuals. To improve the semi-randomized Kaczmarz method further, inspired by Chebyshev’s (weak) law of large numbers, we apply the simple sampling strategy to the semi-randomized Kaczmarz method, and propose a semi-randomized Kaczmarz method with simple random sampling. In the new method, there is no need to calculate probabilities explicitly and it is free of computing residuals of the linear system, and is free of constructing index sets via scanning residuals. Indeed, we only need to compute some elements of residuals corresponding to simple sampling sets, and a small portion of rows of the matrix are utilized. Convergence results are established to show the rationality and feasibility of the two proposed methods. Numerical experiments demonstrate the superiority of the new methods over many state-of-the-art randomized Kaczmarz methods for large-scale linear systems.
Similar content being viewed by others
References
Bai, Z.-Z., Liu, X.-G.: On the Meany inequality with applications to convergence analysis of several row-action iteration methods. Numer. Math. 124, 215–236 (2013)
Bai, Z.-Z., Wu, W.-T.: On convergence rate of the randomized Kaczmarz method. Linear Algebra Appl 553, 252–269 (2018)
Bai, Z.-Z., Wu, W.-T.: On greedy randomized Kaczmarz method for solving large sparse linear systems. SIAM J. Sci. Comput. 40, A592–A606 (2018)
Bai, Z.-Z., Wu, W.-T.: On relaxed greedy randomized Kaczmarz methods for solving large sparse linear systems. Appl. Math. Lett. 83, 21–26 (2018)
Bai, Z.-Z., Wu, W.-T.: On greedy randomized coordinate descent methods for solving large linear least-squares problems. Numer. Linear Algebra Appl. 26, 1–15 (2019)
Bai, Z.-Z., Wu, W.-T.: On partially randomized extended Kaczmarz method for solving large sparse overdetermined inconsistent linear systems. Linear Algebra Appl. 578, 225–250 (2019)
Bai, Z.-Z., Wu, W.-T.: On greedy randomized augmented Kaczmarz method for solving large sparse inconsistent linear systems. SIAM J. Sci. Comput. 43, A3892–A3911 (2021)
Bai, Z.-Z., Wang, L., Muratova, G.V.: On relaxed greedy randomized augmented Kaczmarz methods for solving large sparse inconsistent linear systems. East Asian J. Applied Math. 12, 323–332 (2022)
Borkar, V., Karamchandani, N., Mirani, S.: Randomized Kaczmarz for rank aggregation from pairwise comparisons. In: IEEE Information Theory Workshop (ITW), Cambridge, 2016, pp 389–393 (2016)
Carlton, M.: Probability and statistics for computer scientists. Am. Stat. 62, 271–272 (2008)
Du, K.: Tight upper bounds for the convergence of the randomized extended Kaczmarz and Gauss-Seidel algorithms. Numer. Linear Algebra Appl., Article e2233 (2019)
Du, K., Si, W., Sun, X.: Randomized extended average block Kaczmarz for solving least squares. SIAM J. Sci. Comput. 42, A3541–A3559 (2020)
Du, Y., Hayami, K., Zheng, N., Morikuni, K., Yin, J.: Kaczmarz-type inner-iteration preconditioned flexible GMRES methods for consistent linear systems. SIAM J. Sci. Comput. 43, S345–S366 (2021)
Gordon, R., Bender, R., Herman, G.: Algebraic reconstruction techniques (ART) for three-dimensional electron microscopy and X-ray photography. J. Theor. Biol. 29, 471–481 (1970)
Gower, R., Richtárik, P.: Stochastic dual ascent for solving linear systems, arXiv:1512.06890 (2015)
Gower, R., Richtárik, P.: Randomized iterative methods for linear systems. SIAM J. Matrix Anal. Appl. 36, 1660–1690 (2015)
Gu, C., Liu, Y.: Variant of greedy randomized Kaczmarz for ridge regression. Appl. Numer. Math. 143, 223–246 (2019)
Guo, W., Chen, H., Geng, W., Lei, L.: A modified Kaczmarz algorithm for computerized tomographic image reconstruction. In: IEEE International Conference on Biomedical Engineering and Informatics, pp 1–4 (2009)
Hadgu, A.: An application of ridge regression analysis in the study of syphilis data. Stat. Med. 3, 293–299 (1984)
Hardy, H., Littlewood, J., Pslya, G.: Inequality. Bulletin of the American Mathematical Society, pp. 293–302 (1952)
Hefny, A., Needell, D., Ramdas, A.: Rows versus columns: randomized Kaczmarz or Gauss-Seidel for ridge egression. SIAM J. Sci. Comput. 39, S528–S542 (2017)
Jiang, Y., Wu, G., Jiang, L.: A Kaczmarz method with simple random sampling for solving large linear systems, arXiv:2011.14693 (2020)
Kaczmarz, S.: Approximate solution of systems of linear equations. Int. J. Control. 35, 355–357 (1937)
Lee, S., Kim, H.: Noise properties of reconstructed images in a kilo-voltage on-board imaging system with iterative reconstruction techniques: A phantom study. Phys. Med. 30, 365–373 (2014)
Lei, Y., Zhou, D.: Learning theory of randomized sparse Kaczmarz method. SIAM J. Imaging Sci. 11, 547–574 (2018)
Liu, J., Wright, S.: An accelerated randomized Kaczmarz algorithm. Math. Comput. 85, 153–178 (2016)
Loera, J., Haddock, J., Needell, D.: A sampling Kaczmarz-Motzkin algorithm for linear feasibility. SIAM J. Sci. Comput. 39, S66–S87 (2017)
Necoara, I.: Faster randomized block Kaczmarz algorithms. SIAM J. Matrix Anal. Appl. 40, 1425–1452 (2019)
Ma, A., Needell, D., Ramdas, A.: Convergence properties of the randomized extended Gauss–Seidel and Kaczmarz methods. SIAM J. Matrix Anal. Appl. 36, 1590–1604 (2015)
Needell, D., Srebro, W., Ward, R.: Stochastic gradient descent, weighted sampling, and the randomized Kaczmarz algorithm. Math. Program. 155, 549–573 (2016)
Needell, D., Tropp, J.: Paved with good intentions: Analysis of a randomized block Kaczmarz method. Linear Algebra Appl. 441, 199–221 (2014)
Needell, D., Zhao, R., Zouzias, A.: Randomized block Kaczmarz method with projection for solving least squares. Linear Algebra Appl. 484, 322–343 (2015)
Niu, Y., Zheng, B.: A greedy block Kaczmarz algorithm for solving large-scale linear systems. Appl. Math. Lett., 104, Article 106294 (2020)
Nutini, J., Sepehry, B., Laradji, I., Schmidt, M., Koepke, H., Virani, A.: Convergence rates for greedy Kaczmarz algorithms, and faster randomized Kaczmarz rules using the orthogonality graph. arXiv:1612.07838 (2016)
Ramlau, R., Rosensteiner, M.: An efficient solution to the atmospheric turbulence tomography problem using Kaczmarz iteration. Inverse Probl. 28, 095004 (2012)
Steinerberger, S.: Randomized Kaczmarz converges along small singular vectors. IAM J. Matrix Anal. Appl. 42, 608–615 (2021)
Steinerberger, S.: A weighted randomized Kaczmarz method for solving linear systems. Math. Comput. 90, 2815–2826 (2021)
Strohmer, T., Vershynin, R.: A randomized Kaczmarz algorithm with exponential convergence. J. Fourier Anal. Appl. 15, 262–278 (2009)
Thoppe, G., Borkar, V., Manjunath, D.: A stochastic Kaczmarz algorithm for network tomography. Automatica 50, 910–914 (2014)
van Lith, B., Hansen, P., Hochstenbach, M.: A twin error gauge for Kaczmarz’s iterations. SIAM J. Sci. Comput. 43, S173–S199 (2021)
Wang, C., Ameya, A., Lu, Y.: Randomized Kaczmarz algorithm for inconsistent linear systems: An exact MSE analysis. In: International Conference on Sampling Theory and Applications, Washington DC, pp 498–502 (2015)
Yang, X.: A geometric probability randomized Kaczmarz method for large scale linear systems. Appl. Numer. Math. 164, 139–160 (2021)
Zhang, J.: A new greedy Kaczmarz algorithm for the solution of very large linear systems. Appl. Math. Lett. 91, 207–212 (2019)
Zhang, Y., Li, H.: Greedy Motzkin-Kaczmarz methods for solving linear systems. Numer. Linear Algebra Appl. https://doi.org/10.1002/nla.2429 (2021)
Zouzias, A., Freris, N.: Randomized extended Kaczmarz for solving least squares. SIAM J. Matrix Anal. Appl. 34, 773–793 (2013)
Acknowledgements
Special thanks to the two anonymous reviewers for their insightful comments and invaluable suggestions that greatly improved the representation of this paper. Meanwhile, we are grateful to Prof. Zhong-Zhi Bai, Mr. Shunchang Li, and Dr. Bo Feng for helpful discussions on an early version of this paper.
Funding
Gang Wu is supported by the National Natural Science Foundation of China under grant 12271518, the Key Research and Development Project of Xuzhou Natural Science Foundation under grant KC22288, and the Open Project of Key Laboratory of Data Science and Intelligence Education of the Ministry of Education under grant DSIE202203.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare no competing interests.
Additional information
Communicated by: Lothar Reichel
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Jiang, Y., Wu, G. & Jiang, L. A semi-randomized Kaczmarz method with simple random sampling for large-scale linear systems. Adv Comput Math 49, 20 (2023). https://doi.org/10.1007/s10444-023-10018-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s10444-023-10018-2
Keywords
- Randomized Kaczmarz method (RK)
- Greedy randomized Kaczmarz method (GRK)
- Large-scale linear system
- Simple random sampling
- Relative homogeneous residual
- Chebyshev’s (weak) law of large numbers