It seems like there is no FFN in current implementation.

StatsDLMathsRecomSys / Inductive-representation-learning-on-temporal-graphs

297 stars 62 forks source link

It seems like there is no FFN in current implementation. #11

Open baoleai opened 2 years ago

baoleai commented 2 years ago

Hi, it seems like there is no FFN in current implementation. According to TGAT paper, the attention output h should be concatenated with the target node feature, but I don't find that in current version.