You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I looked through the sequence model, it looks like the model is trying to learn a user representation from the given sequence alone: user_representation, _ = self._net.user_representation( sequence_var )
It does not involve who is that specific user for that sequence. Look like to_sequence() in Interactions class includes both sequences and sequence_users, so it could potentially know which user generates that sequence, but the model does not use it anyway.
If my code reading is correct, the above seems like the logic of the model. My question is, for example if one user has interaction like [A, B, C], the other user is [A, B, D], and another user is [A, B, E], will the model tend to embed C, D and E in a close way to reduce the cost? Does the model ignore the difference of users profiles themselves in this case?
The text was updated successfully, but these errors were encountered:
I looked through the sequence model, it looks like the model is trying to learn a user representation from the given sequence alone:
user_representation, _ = self._net.user_representation( sequence_var )
It does not involve who is that specific user for that sequence. Look like to_sequence() in Interactions class includes both sequences and sequence_users, so it could potentially know which user generates that sequence, but the model does not use it anyway.
If my code reading is correct, the above seems like the logic of the model. My question is, for example if one user has interaction like [A, B, C], the other user is [A, B, D], and another user is [A, B, E], will the model tend to embed C, D and E in a close way to reduce the cost? Does the model ignore the difference of users profiles themselves in this case?
The text was updated successfully, but these errors were encountered: