Collaborative Ranking for Local Preferences
Berk Kapicioglu, David S. Rosenberg, Robert E. Schapire, Tony Jebara · 2014
For many collaborative ranking tasks, we have access to relative preferences among subsets of items, but not to global preferences among all items. To address this, we intro-duce a matrix factorization framework called Collaborative Local Ranking (CLR). We jus-tify CLR by proving a bound on its gener-alization error, the first such bound for col-laborative ranking that we know of. We then derive a simple alternating minimization al-gorithm and prove that its running time is independent of the number of training exam-ples. We apply CLR to a novel venue recom-mendation task and demonstrate that it out-performs state-of-the-art collaborative rank-ing methods on real-world data sets. 1