Reference implementation (example) of the model proposed in the paper:

Diffusion Improves Graph Learning
by Johannes Klicpera, Stefan Weißenberger, Stephan Günnemann
Published at NeurIPS 2019.

This repository primarily contains a demonstration of enhancing a graph convolutional network (GCN) with graph diffusion convolution (GDC) in the notebook gdc_demo.ipynb.

The repository uses these packages:

pyyaml
tqdm>=4.36
numpy
scipy
seaborn
pytorch>=1.3
pytorch_geometric

GDC is also implemented as a transformation (preprocessing step) in PyTorch Geometric. So you can just apply it to your own dataset and see how your existing PyG model improves!

Please contact [email protected] in case you have any questions.

Please cite our paper if you use the model or this code in your own work:

@inproceedings{klicpera_diffusion_2019,
  title = {Diffusion Improves Graph Learning},
  author = {Klicpera, Johannes and Wei{\ss}enberger, Stefan and G{\"u}nnemann, Stephan},
  booktitle={Conference on Neural Information Processing Systems (NeurIPS)},
  year = {2019}
}