Matrix Factorization via Deep Learning
This work solves matrix completion issues for applications like recommendation systems, but it is incremental as it builds on existing deep learning methods.
The paper tackles the problem of matrix completion by addressing two drawbacks of deep-learning-based models: inability to handle unseen rows/columns and degraded performance for discrete predictions, proposing a deep matrix factorization model with joint training of factorization and discretization, and shows efficacy on a real movie rating dataset.
Matrix completion is one of the key problems in signal processing and machine learning. In recent years, deep-learning-based models have achieved state-of-the-art results in matrix completion. Nevertheless, they suffer from two drawbacks: (i) they can not be extended easily to rows or columns unseen during training; and (ii) their results are often degraded in case discrete predictions are required. This paper addresses these two drawbacks by presenting a deep matrix factorization model and a generic method to allow joint training of the factorization model and the discretization operator. Experiments on a real movie rating dataset show the efficacy of the proposed models.