HGT self-supervised training for node embedding

Models & Apps
#1

Hi DGL staff and community,

I’m looking to adapt this example to a self-supervised task - training node embedding for one type of node in my heterogenous graph based on some supervision signal from the graph structure. Could you please advise where to start? Thanks.

#2

trying to combine HGT with Deep Graph Infomax as the start point. Any suggestions would be great! Thanks

#3

Issue resolved - I figured it out. Deep graph infomax is a nice framework. In order to make it work for heterogenous graph, it looks like we just need to use an encoder like HGT that works for heterogenous graph.

#4

Sorry for the late response, and I am glad you have your issue solved, let us know if you have further questions.

1 Like