Saved in:
Bibliographic Details
Main Authors: He, Huan, Tang, Ziyuan, Zhao, Shifan, Saad, Yousef, Xi, Yuanzhe
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2306.00325
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914734890024960
author He, Huan
Tang, Ziyuan
Zhao, Shifan
Saad, Yousef
Xi, Yuanzhe
author_facet He, Huan
Tang, Ziyuan
Zhao, Shifan
Saad, Yousef
Xi, Yuanzhe
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2306_00325
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle NLTGCR: A class of Nonlinear Acceleration Procedures based on Conjugate Residuals
He, Huan
Tang, Ziyuan
Zhao, Shifan
Saad, Yousef
Xi, Yuanzhe
Numerical Analysis
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.
title NLTGCR: A class of Nonlinear Acceleration Procedures based on Conjugate Residuals
topic Numerical Analysis
url https://arxiv.org/abs/2306.00325