Learning to Rank for Collaborative Filtering

Jean-François Pessiot, Vinh Truong, Nicolas Usunier, Massih-Reza Amini, Patrick Gallinari
Laboratoire d'Informatique Paris 6
104, avénue du Président Kennedy
75016 Paris

This paper presents the design of a new machine learning based recommendation engine. The aim of this system is to generate real-time personalized recommendations. We present a Collaborative Filtering (CF) approach to this problem which consists to make automatic predictions (filtering) about the interests of a user by collecting taste information from many other users (collaborating). Though the principle is simple, the major diffito come out with a really efficient algorithm capable to handle huge volume in a real time basis is very complex. As a result, collaborative approaches involve two major constraints.