PINNS – Research Code Repository

This repository gathers the codes developed during the PhD thesis of Hugo Boulenc. The work focuses on Physics-Informed Neural Networks (PINNs) and related machine learning architectures applied to inverse problems for flood dynamics.

The thesis manuscript (in French) is available at the following link:

Thesis manuscript

Part of this work led to a scientific publication in the journal Inverse Problems, available at the following links:

Journal article
Preprint

Repository structure

PINNS/
│
├── code/              # Source codes for PINNs and DeepONets
├── data/              # Datasets required to reproduce the experiments
├── trained_models/    # Some pre-trained models
└── README.md

Datasets and trained models

Due to their large size, the datasets and trained models are not included directly in this repository. They can be downloaded using the following links:

Datasets & Trained models

Installation of datasets and models

After downloading the archives:

1. Extract the downloaded files.
2. Replace the corresponding folders in the repository.

Expected final structure:

PINNS/
│
├── code/
├── data/              <-- replace with downloaded data folder
├── trained_models/    <-- replace with downloaded trained_models folder
└── README.md

Purpose of this repository

This repository aims to:

• provide the implementation of the methods developed during the PhD,
• allow reproduction of the experiments,
• make the research code publicly available.

How to cite this work

Boulenc, H., Bouclier, R., Garambois, P. A., & Monnier, J. (2025). Spatially-distributed parameter identification by physics-informed neural networks illustrated on the 2D shallow-water equations. Inverse Problems, 41(3), 035006.
DOI: https://doi.org/10.1088/1361-6420/adb0e7

BibTeX entry:

@article{boulenc2025spatially,
  title={Spatially-distributed parameter identification by physics-informed neural networks illustrated on the 2D shallow-water equations},
  author={Boulenc, Hugo and Bouclier, Robin and Garambois, Pierre-Andr{\'e} and Monnier, Jerome},
  journal={Inverse Problems},
  volume={41},
  number={3},
  pages={035006},
  year={2025},
  publisher={IOP Publishing}
}