Saved in:
Bibliographic Details
Main Authors: Wang, Zhen, Liu, Yun, Cui, Chen, Shu, Shi
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2412.08076
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915058439684096
author Wang, Zhen
Liu, Yun
Cui, Chen
Shu, Shi
author_facet Wang, Zhen
Liu, Yun
Cui, Chen
Shu, Shi
contents Recently, designing neural solvers for large-scale linear systems of equations has emerged as a promising approach in scientific and engineering computing. This paper first introduce the Richardson(m) neural solver by employing a meta network to predict the weights of the long-step Richardson iterative method. Next, by incorporating momentum and preconditioning techniques, we further enhance convergence. Numerical experiments on anisotropic second-order elliptic equations demonstrate that these new solvers achieve faster convergence and lower computational complexity compared to both the Chebyshev iterative method with optimal weights and the Chebyshev semi-iteration method. To address the strong dependence of the aforementioned single-level neural solvers on PDE parameters and grid size, we integrate them with two multilevel neural solvers developed in recent years. Using alternating optimization techniques, we construct Richardson(m)-FNS for anisotropic equations and NAG-Richardson(m)-WANS for the Helmholtz equation. Numerical experiments show that these two multilevel neural solvers effectively overcome the drawback of single-level methods, providing better robustness and computational efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2412_08076
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Momentum-Accelerated Richardson(m) and Their Multilevel Neural Solvers
Wang, Zhen
Liu, Yun
Cui, Chen
Shu, Shi
Numerical Analysis
Recently, designing neural solvers for large-scale linear systems of equations has emerged as a promising approach in scientific and engineering computing. This paper first introduce the Richardson(m) neural solver by employing a meta network to predict the weights of the long-step Richardson iterative method. Next, by incorporating momentum and preconditioning techniques, we further enhance convergence. Numerical experiments on anisotropic second-order elliptic equations demonstrate that these new solvers achieve faster convergence and lower computational complexity compared to both the Chebyshev iterative method with optimal weights and the Chebyshev semi-iteration method. To address the strong dependence of the aforementioned single-level neural solvers on PDE parameters and grid size, we integrate them with two multilevel neural solvers developed in recent years. Using alternating optimization techniques, we construct Richardson(m)-FNS for anisotropic equations and NAG-Richardson(m)-WANS for the Helmholtz equation. Numerical experiments show that these two multilevel neural solvers effectively overcome the drawback of single-level methods, providing better robustness and computational efficiency.
title Momentum-Accelerated Richardson(m) and Their Multilevel Neural Solvers
topic Numerical Analysis
url https://arxiv.org/abs/2412.08076