Paper abstract

Improving Maximum Margin Matrix Factorization

Markus Weimer - Technische Universitat Darmstadt, Germany
Alexandros Karatzoglou - INSA de Rouen, France
Alex Smola - NICTA, Australia

Session: Matrix Factorization
Springer Link: http://dx.doi.org/10.1007/978-3-540-87479-9_12

Collaborative filtering is a popular method for personalizing product recommendations. Maximum Margin Matrix Factorization (MMMF) has been proposed as one successful learning approach to this task and has been recently extended to structured ranking losses. In this paper we discuss a number of extensions to MMMF by introducing offset terms, item dependent regularization and a graph kernel on the recommender graph. We show equivalence between graph kernels and the recent MMMF extensions by Mnih and Salakhutdinov. Experimental evaluation of the introduced extensions showimproved performance over the original MMMF formulation.