A Tensor-flow implementation of a Recommender System which takes in a sparse user-item rating matrix to generate new entries using which items can be recommended to users.
It is inspired by the model described in the publication: Deep Variational Autoencoder with Shallow Parallel Path for Top-N Recommendation (VASP) It is recommended to read the above paper for better unerstanding of the implementation.
Note: A closed-form solution of EASE is implemented here instead of the Neural EASE model.
Uthpal P Siddhanth M Sathvik Rao
A sparse matrix containing the ratings of movies rated by different users obtained from MovieLens100k.
A Hadamard product of the matrix generated using the EASE Model and the matrix generated using a Varational Auto-Encoder. This matrix contains new user-movie rating entries which is used to provide new movie recommendations to the users.