How to extract node embedding from GCN model?

Questions
#1

I’m new in graphs (and dgl) and I was going through the pytorch examples. There for GraphSAGE we have inference function to extract node embedding but for GCN I didn’t find any such function. I want to extract node embedding from GCN model after semi-supervised training. Can you please guide me?
Any help will be highly appreciated.

#2

The GraphSAGE example performs sampling-based training and there is an inference function for extracting the embedding of all nodes. The GCN example does not perform sampling-based training and with model.eval(), the output of the forward function is the embedding of all nodes.

#3

Hi @mufeili ,

Just to make things clear I wanna ask that at line dgl/train.py at 411bef5498bc79641858ea15ee79b6085c855187 · dmlc/dgl · GitHub we are calculating embedding?

Thanks

#4

That’s correct. [placeholder for 20 characters]

#5

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.