Published February 18, 2026 | Version v1
Software Open

Preconditioning and Numerical Stability in Neural Network Training for Parametric PDEs (Python code)

Authors/Creators

  • 1. Institut für Geometrie und Praktische Mathematik, RWTH Aachen University
  • 2. Mathematics Department, University of South Carolina

Description

This repository contains a PyTorch implementation for the paper Preconditioning and Numerical Stability in Neural Network Training for Parametric PDEs, by Markus Bachmayr, Wolfgang Dahmen, Chenguang Duan, and Mathias Oster.

 

Files

clean.zip

Files (28.6 kB)

Name Size Download all
md5:90529e19b74a5828220f8c4ca0b30745
28.6 kB Preview Download

Additional details

Related works

Is supplement to
Preprint: arXiv:2601.23185 (arXiv)

Funding

Deutsche Forschungsgemeinschaft
SFB 1481 442047500
European Research Council
COCOA 101170147
U.S. National Science Foundation
FRG DMS-2245097
U.S. National Science Foundation
DMS-2513112
U.S. National Science Foundation
RTG DMS-2038080

Dates

Available
2026-02

Software

Programming language
Python