Closed zolekode closed 4 years ago
Hi there,
A large number of those methods is transductive - which means that you could not make a consistent API and for certain classes, the predict / score methods should not return anything.
E.g. from the implemented methods only MUSAE is inductive.
Hi, thanks a lot this helps deffo
Hi, for attribute based approaches, is there a way to get embeddings at test time on unseen data ? For example: after
embedder.fit(graph, X)
, can i input for exampleembbeder.predict([some unseen input features])
to get embeddings at test time ?