search

pyg-team / pytorch_geometric

Graph Neural Network Library for PyTorch
https://pyg.org
MIT License
20.85k stars 3.6k forks source link

How to train a heterogeneous graph by using unsupervised ? #3159

Open Kang9779 opened 2 years ago

Kang9779 commented 2 years ago

I have a question about, How can we train a heterogeneous graph by using unsupervised ?

rusty1s commented 2 years ago

This is an interesting question. Most self-supervised/unsupervised graph machine learning papers only consider homogeneous graphs, and it is up to the user to lift them to the heterogeneous case. One approach I can think of is the MetaPath2Vec algorithm implemented in PyG. If you have any additional references about unsupervised heterogeneous graph learning, please let me know.

One interesting direction would be to lift the examples/infomax.py example to heterogeneous graphs as well, but which might require some modifications. Let me know if you are interested in working on this together.