mims-harvard/decagon

Confusion about weights update

colivarese opened this issue · 0 comments

It seems that the model takes into consideration a single edge type on each batch iteration, so it only uses that corresponding W_r, is only that matrix updated on that iteration?