Closed shivamtundele closed 2 years ago
Thanks for the asking! Now we can build user and item features into the model by domain adoptation. This technique is archived by adding to the loss the distance between the projection from that user/item feature to the embedding space and the embedding of the user/item.
For details, see the README on "adoptors" and section 3.3 of the paper.
Thank you! I would love to contribute on this repo to do that. Are there any contribution guidelines?
Hi! Is there a way to add user and item features to the model just like LightFM?