The goal of a recommendation system is to predict the interest of a user in a
given item by exploiting the existing set of ratings as well as certain
user/item features. A standard approach to modeling this problem is Inductive
Matrix Completion where the predicted rating is modeled as