Probabilistic Matrix Factorization

Abstract

1 Introduction One of the most popular approaches to collaborative filtering is based on low-dimensional factormodels. The idea behind such models is that attitudes or preferences of a user are determined by a small number of unobserved factors. In a linear factor model, a user's preferences are modeledby linearly combining item factor vectors using user-specific coefficients. For example, for N usersand M movies, the N * M preference matrix R is given by the product of an N * D user coefficientmatrix