Open luiscastro193 opened 5 years ago
Hey @luiscastro193 -- great question!
Implementing an autoencoder based recommender would be possible once I've finished https://github.com/jfkirk/tensorrec/issues/86
Once we have this working, a user's previous interactions could be fed in as features and autoencoder shaped nets could consume this information to make predictions and learn representations.
I'm going to leave this open as a reminder to create an example of this once #86 is finished.
I came across the same issue as #93 and #116. Investigating further, I found this really interesting approach which would also solve that typical use case: https://arxiv.org/abs/1901.00415
Would you consider adding an autoencoder model?