Open demien opened 5 years ago
NFM uses the feature embeddings of FM to init the embedding layer. The deep layers are learned from random init.
On Mon, Nov 26, 2018 at 4:29 PM demien notifications@github.com wrote:
As the code it seems do not load the deep layer params if use the pretrain model.
https://github.com/hexiangnan/neural_factorization_machine/blob/master/NeuralFM.py#L202-L217
Is that on purpose which means daily model has a brand new deep layer params?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/hexiangnan/neural_factorization_machine/issues/6, or mute the thread https://github.com/notifications/unsubscribe-auth/ABGxjrLqxNioq9jEV0zVeY0h6FKWAXfzks5uy6ZmgaJpZM4Yyiyn .
As the code it seems do not load the deep layer params if use the pretrain model.
https://github.com/hexiangnan/neural_factorization_machine/blob/master/NeuralFM.py#L202-L217
Is that on purpose which means daily model has a brand new deep layer params?