Hyperbolic Graph Convolutional Networks in PyTorch

This repository is a fork of https://github.com/HazyResearch/hgcn, and additions/modifications are made by Eli and Chris.

We use their implementation of Hyperbolic Graph Convolutions [1] in PyTorch to examine how embedding on different manifolds can impact performance on link prediction and also node classification.

See examples in this Colab.

This is also a class project for CS468 at Stanford.

Most of the code was forked from the following repositories

• hgcn
• pygcn
• gae
• hyperbolic-image-embeddings
• pyGAT
• poincare-embeddings
• geoopt

References

[1] Chami, I., Ying, R., Ré, C. and Leskovec, J. Hyperbolic Graph Convolutional Neural Networks. NIPS 2019.

[2] Nickel, M. and Kiela, D. Poincaré embeddings for learning hierarchical representations. NIPS 2017.

[3] Ganea, O., Bécigneul, G. and Hofmann, T. Hyperbolic neural networks. NIPS 2017.

[4] Kipf, T.N. and Welling, M. Semi-supervised classification with graph convolutional networks. ICLR 2017.

[5] Veličković, P., Cucurull, G., Casanova, A., Romero, A., Lio, P. and Bengio, Y. Graph attention networks. ICLR 2018.