Supplementary material for the following paper: Barwey, Shivam, Varun Shankar, Venkatasubramanian Viswanathan, and Romit Maulik. "Multiscale Graph Neural Network Autoencoders for Interpretable Scientific Machine Learning." Journal of Computational Physics (2023). Contains example code that trains GNN on unstructured MNIST dataset, as demonstrated in Appendix A in the paper.
Paper: https://www.sciencedirect.com/science/article/pii/S0021999123006320
Please cite using the following BibTeX entry upon usage or reference of any code/data:
@article{BARWEY2023112537,
title = {Multiscale Graph Neural Network Autoencoders for Interpretable Scientific Machine Learning},
journal = {Journal of Computational Physics},
pages = {112537},
year = {2023},
issn = {0021-9991},
doi = {https://doi.org/10.1016/j.jcp.2023.112537},
url = {https://www.sciencedirect.com/science/article/pii/S0021999123006320},
author = {Shivam Barwey and Varun Shankar and Venkatasubramanian Viswanathan and Romit Maulik}
}
dataset: unstructured MNIST dataset.
main.py: example code for graph neural network training.
models.py: class definitions for graph neural network models and layers.
pooling.py: contains pooling operations used in GNN architectures