Back to Search
Start Over
NLTGCR: A class of Nonlinear Acceleration Procedures based on Conjugate Residuals
- Source :
- SIAM Journal on Matrix Analysis and Applications, Volume 45, Issue 1, pp. 1-827 (2024)
- Publication Year :
- 2023
-
Abstract
- This paper develops a new class of nonlinear acceleration algorithms based on extending conjugate residual-type procedures from linear to nonlinear equations. The main algorithm has strong similarities with Anderson acceleration as well as with inexact Newton methods - depending on which variant is implemented. We prove theoretically and verify experimentally, on a variety of problems from simulation experiments to deep learning applications, that our method is a powerful accelerated iterative algorithm.
- Subjects :
- Mathematics - Numerical Analysis
Subjects
Details
- Database :
- arXiv
- Journal :
- SIAM Journal on Matrix Analysis and Applications, Volume 45, Issue 1, pp. 1-827 (2024)
- Publication Type :
- Report
- Accession number :
- edsarx.2306.00325
- Document Type :
- Working Paper
- Full Text :
- https://doi.org/10.1137/23M1576360