search

pyg-team / pytorch_geometric

Graph Neural Network Library for PyTorch
https://pyg.org
MIT License
21.41k stars 3.67k forks source link

Reimplementation of a graph-based recommender system SR-GNN? #200

Open RuihongQiu opened 5 years ago

RuihongQiu commented 5 years ago

❓ Questions & Help

I have used PyTorch_Geometric to reimplement SR-GNN in this repo SR-GNN_PyTorch-Geometric of the paper Session-based Recommendation with Graph Neural Networks.

If possible, is there any chance to help promote this repo as an example to build a recommender system with this great graph neural network package?

@rusty1s

RuihongQiu commented 5 years ago

I have finished the reimplementation in my SR-GNN_PyTorch-Geometric repo and achieve the similar result with the paper.

Welcome pull request!

rusty1s commented 5 years ago

Really cool :) Would love to see a pull request on this one if you have time.

RuihongQiu commented 5 years ago

But the full code includes a folder and it seems too heavy to add into the examples folder in your code.

rusty1s commented 5 years ago

I do not think there is a need to include the raw dataset files into PyG. Just download them on the fly like we do it with all datasets.

RuihongQiu commented 5 years ago

Yeah. I have 4 .py files. And do I need to create a folder inside examples to put them in?

rusty1s commented 5 years ago

I suggest adding the datasets to the torch_geometric.datasets subpackage, and the model to the torch_geometric.nn.models subpackage. You can then create a single-file example to train the model (e.g., like infomax.py or autoencoder.py). What do you think?

RuihongQiu commented 5 years ago

Hmm. Good idea and I will have a try : )

Jeriousman commented 1 year ago

Did you make it happen @RuihongQiu? If so, where can I find your work?