{"id":"arxiv/2610.16542","url":"https://arxiv.org/abs/2610.16542","tags":["math.NA","cs.LG"],"score":0,"title":"Neural Network Preconditioning: A Case Study for the Solution of the Parametric Helmholtz Equation","domain":"math.NA","source":"arxiv","authors":["Luc Giraud","Carola Kruse","Paul Mycek","Maksym Shpakovych","Yanfei Xiang"],"reviews":0,"subject":"math.NA","summary":"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 an 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…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-12","reactions":{}}