Link prediction not only for knowledge graphs

Has anybody done link prediction with DGL for other datasets besides knowledge graphs? If not, any ideas on how to repurpose the R-GCN for other types of datasets? I’m interested in matrix completion/recommender systems also for homogeneous graphs, e.g. friendships on social networks


We have GCMC, which is on Movielens dataset.

I’m not sure about homogeneous graph. Is there any dataset to do so?

Same question. Thank you.