Authors : Xavier Cassagnou^1,2, Aurélien Casagrandi^1,4, Elodie Noëlé^1,3, Christophe Millet^1,2, Mathilde Mougeot^2,4

Affiliation :

¹CEA DAM DIF

²Centre Borelli, ENS Paris-Saclay, Université Paris-Saclay

³DGA

⁴ENSIIE

Corresponding author: christophe.millet@cea.fr

Link to the article: X. Cassagnou et al., Dynamic graph neural networks for seismic characterization (To be published)

Date : 19th June 2025

License : MIT

---

Run on Google Colab

You can try out the code and materials using the following resources:

• ▶️ Open in Google Colab
• 📁 Download Materials from Google Drive

Due to the large size of the training/testing data and models related to oversmoothing (over 10 GB), they are not included directly in the GitHub repository. However, you can access them via the provided Google Drive links.

---

Repository Content

Folder/File	Description
modules/	Source code for the results'plots, models and utilities.
oversmoothing/	Source code and plots for the oversmoothing part.
data/	Sample synthetic or preprocessed datasets.
images/	Images
README.md	This file.
requirements.txt	List of mandatory Python modules for running this tutorial.
License	License information.
GNN_Tutorial.ipynb	Jupyter notebook for launching the tutorial

---

Visual Overview

Animated Output (Training or Results)

---

License

MIT

---