Neural Network Preconditioning: A Case Study for the Solution of the Parametric Helmholtz Equation
Abstract
This work presents a hybrid numerical approach for solving linear systems arising from the discretization of the two‐dimensional parametric Helmholtz equation. A convolutional neural network based on the U‐Net architecture is trained in a self‐supervised manner (i.e., without explicitly generating reference solutions) to approximate the inverse of the discretized Helmholtz operator, using a loss function involving the residual norm of the linear system. The trained network is used as a nonlinear preconditioner within the flexible GMRES (FGMRES) algorithm. Numerical experiments show that while the neural network is not accurate enough to act as a standalone solver, it significantly improves the convergence of FGMRES when employed as a preconditioner. The neural preconditioner demonstrates robust performance and generalization capabilities with respect to variations in the velocity field and the domain size. Comparisons with algebraic preconditioners, based on LU factorizations of approximations of the original system matrices, indicate the superior efficiency of the neural approach under equivalent conditions. We believe that the proposed method is not tied to a specific neural architecture and can be extended to other parametric PDEs.
Versions
Related
- Quasi-Helmholtz Calderón Multiplicative Preconditioning for Higher-Order Global Multi-Trace Integral Equations
- On solutions of the modified Helmholtz equation in a planar domain
- Cartesian PML truncation for the Helmholtz equation is exponentially accurate at high frequency
- Local rigidity of conformally Euclidean metrics for the anisotropic Helmholtz equation at fixed energy
- Increasing stability for the inverse potential problem of the fractional Helmholtz equation
BibTeX
@article{giraud2026neural,
title = {{Neural Network Preconditioning: A Case Study for the Solution of the Parametric Helmholtz Equation}},
author = {Luc Giraud and Carola Kruse and Paul Mycek and Maksym Shpakovych and Yanfei Xiang},
journal = {Numerical Linear Algebra with Applications},
year = {2026},
month = oct,
doi = {10.1002/nla.70123},
url = {https://hal.science/hal-05794192v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Neural Network Preconditioning: A Case Study for the Solution of the Parametric Helmholtz Equation
AU - Luc Giraud
AU - Carola Kruse
AU - Paul Mycek
AU - Maksym Shpakovych
AU - Yanfei Xiang
PY - 2026
DA - 2026/10/09
JO - Numerical Linear Algebra with Applications
DO - 10.1002/nla.70123
UR - https://hal.science/hal-05794192v1
AB - This work presents a hybrid numerical approach for solving linear systems arising from the discretization of the two‐dimensional parametric Helmholtz equation. A convolutional neural network based on the U‐Net architecture is trained in a self‐supervised manner (i.e., without explicitly generating reference solutions) to approximate the inverse of the discretized Helmholtz operator, using a loss function involving the residual norm of the linear system. The trained network is used as a nonlinear preconditioner within the flexible GMRES (FGMRES) algorithm. Numerical experiments show that while the neural network is not accurate enough to act as a standalone solver, it significantly improves the convergence of FGMRES when employed as a preconditioner. The neural preconditioner demonstrates robust performance and generalization capabilities with respect to variations in the velocity field and the domain size. Comparisons with algebraic preconditioners, based on LU factorizations of approximations of the original system matrices, indicate the superior efficiency of the neural approach under equivalent conditions. We believe that the proposed method is not tied to a specific neural architecture and can be extended to other parametric PDEs.
SN - 1070-5325
SN - 1099-1506
ER -