Geometric Deep Learning models

This example shows how to use geometric deep learning models defined in dgl.nn.pytorch.conv for graph classification.

Currently we support following models:

Image Classification on MNIST

By transforming images to graphs, graph classifcation algorithms could be applied to image classification problems.

python mnist.py --model cheb --gpu 0
python mnist.py --model monet --gpu 0

We thank Xavier Bresson for providing code for graph coarsening algorithm and grid graph building in
CE7454_2019 Labs.