This is a repository for a model from Embarrassingly Shallow Autoencoders for Sparse Data.

This model is cool because it is a closed-form solution.

Similiarity matrix is calculated on CPU with numpy.

pandas.DataFrame with columns user_id and item_id both for fit and predict.

It may also use ratings from column rating if implicit parameter is set to False.

pandas.DataFrame with columns user_id, item_id, score