NLTGCR: A class of Nonlinear Acceleration Procedures based on Conjugate Residuals
Numerical Analysis
2024-04-02 v3 Numerical Analysis
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.
Keywords
Cite
@article{arxiv.2306.00325,
title = {NLTGCR: A class of Nonlinear Acceleration Procedures based on Conjugate Residuals},
author = {Huan He and Ziyuan Tang and Shifan Zhao and Yousef Saad and Yuanzhe Xi},
journal= {arXiv preprint arXiv:2306.00325},
year = {2024}
}