my silly question - Githubissues

INK-USC / temporal-gcn-lstm

Code for Characterizing and Forecasting User Engagement with In-App Action Graphs: A Case Study of Snapchat

77 stars 26 forks source link

my silly question #2

Closed Bitliuliu closed 4 years ago

Bitliuliu commented 4 years ago

I'm curious why networkx is already used and dgl is added

yozenliu commented 4 years ago

Hey! DGL framework supports message passing which allows us to aggregate neighbor representations and do convolution easily. Our input graphs are in networkx format, but you can of course directly load your graphs into DGLGraph objects!

Bitliuliu commented 4 years ago

Thank you for your reply. My recent project involves the joint training of PYG graph neural network and lstm network. I do n’t know if I can discuss it with you.

yozenliu commented 4 years ago

Hey sorry I missed this one, you can shoot me an email if it's still relevant!