Abstract
For systems of nonlinear equations, we propose a new version of the Gauss–Newton method based on the idea of using an upper bound for the residual norm of the system and a quadratic regularization term. The global convergence of the method is proved. Under natural assumptions, global linear convergence is established. The method uses an adaptive strategy to choose hyperparameters of a local model, thus forming a flexible and convenient algorithm that can be implemented using standard convex optimization techniques.
Similar content being viewed by others
REFERENCES
A. I. Golikov, Yu. G. Evtushenko, and I. E. Kaporin, Comput. Math. Math. Phys. 59 (12), 2017–2032 (2019). https://doi.org/10.1134/S0044466919120093.
Yu. Nesterov, Lectures on Convex Optimization (Springer, Berlin, 2018).
A. V. Gasnikov, Modern Numerical Optimization Methods: Universal Gradient Descent (Mosk. Tsentr Neprer. Mat. Obrazovan., Moscow, 2020) [in Russian].
Yu. Nesterov, Optim. Methods Software 22 (3), 469–483 (2007).
E. Gorbunov, F. Hanzely, and P. Richtárik, “A unified theory of SGD: Variance reduction, sampling, quantization and coordinate descent,” International Conference on Artificial Intelligence and Statistics, Proceedings of Machine Learning Research (2020), pp. 680–690.
B. T. Polyak, USSR Comput. Math. Math. Phys. 3 (4), 864–878 (1963).
Yu. Nesterov, “Flexible modification of Gauss–Newton method,” CORE Discussion Papers (2021).
N. Yudin and A. Gasnikov, Preprint No. 2813 (WIAS, 2021). https://doi.org/10.20347/wias.preprint.2813
Funding
This work was supported by the Russian Science Foundation, project no. 21-71-30005.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
The author declares that he has no conflicts of interest.
Additional information
Translated by I. Ruzanova
Rights and permissions
About this article
Cite this article
Yudin, N.E. Adaptive Gauss–Newton Method for Solving Systems of Nonlinear Equations. Dokl. Math. 104, 293–296 (2021). https://doi.org/10.1134/S1064562421050161
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1134/S1064562421050161