Analysis of Variational Bayesian Matrix Factorization

Abstract. Recently, the variational Bayesian approximation was applied to probabilistic matrix factorization and shown to perform very well in experiments. However, its good performance was not completely understood beyond its experimental success. The purpose of this paper is to theoretically elucidate properties of a variational Bayesian matrix factorization method. In particular, its mechanism of avoiding overfitting is analyzed. Our analysis relies on the key fact that the matrix factorization model induces non-identifiability, i.e., the mapping between factorized matrices and the original matrix is not one-to-one. The positivepart James-Stein shrinkage operator and the Marcenko-Pastur law—the limiting distribution of eigenvalues of the central Wishart distribution—play important roles in our analysis.

• Type: Conference
• Year: 2009
• Where: PAKDD
• Authors: Shinichi Nakajima, Masashi Sugiyama
• Keywords: Bayesian Matrix Factorization, Data Mining, Matrix Factorization, PAKDD 2009, Probabilistic Matrix Factorization