Improving Maximum Margin Matrix Factorization
Markus Weimer, Alexandros Karatzoglou and Alexander J. Smola
Test of Time Award at ECML/PKDD 2018
Best machine learning paper @ ECML/PKDD 2008
Abstract
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 (Advances in Neural Information Processing Systems 20, 2008). Experimental evaluation of the introduced extensions show improved performance over the original MMMF formulation.
BibTeX
The paper was published as an abstract in the proceedings of ECML:
@inproceedings{Weimer:2008lq,
author = {Markus Weimer and Alexandros Karatzoglou and Alexander J. Smola},
Booktitle = {Machine Learning and Knowledge Discovery in Databases},
Editor = {Walter Daelemans and Bart Goethals and Katharina Morik},
Isbn = {978-3-540-87478-2},
Pages = {14--14},
Series = {LNAI},
Title = {Improving Maximum Margin Matrix Factorization},
Volume = {5211},
Year = {2008},
}
And in the Machine Learning Journal
@article{Weimer:2008zl,
Author = {Markus Weimer and Alexandros Karatzoglou and Alexander J. Smola},
Journal = {Machine Learning},
Number = {3},
Pages = {263--276},
Title = {Improving maximum margin matrix factorization},
Volume = {72},
Year = {2008},
}