search

dmlc / dgl

Python package built to ease deep learning on graph, on top of existing DL frameworks.
http://dgl.ai
Apache License 2.0
13.51k stars 3.02k forks source link

Transductive or inductive learning on OGB Products? #1513

Closed Elizabeth1997 closed 4 years ago

Elizabeth1997 commented 4 years ago

❓ Questions and Help

Hi dgl team, I would like to know on OGB products dataset, you used transductive or inductive learning to get the result posted on the LeaderBoard? Thank you for answering in advance!

BarclayII commented 4 years ago

Good question!

I used the transductive learning scenario, and I think it matches the setting of ClusterGCN and Full-batch GraphSAGE OGB has provided in their examples. However, it looks like OGB's GraphSAINT example works under an inductive learning setting.

I have raised an issue https://github.com/snap-stanford/ogb/issues/27 in OGB as well.

Elizabeth1997 commented 4 years ago

I see, got it. Thanks for your reply! Yes, exactly, as you said, only GraphSAINT example works under an inductive learning setting.