Nonparametric factor analysis with beta process priors

We propose a nonparametric extension to the factor analysis problem using a beta process prior. This beta process factor analysis (BPFA) model allows for a dataset to be decomposed into a linear combination of a sparse set of factors, providing information on the underlying structure of the observations. As with the Dirichlet process, the beta process is a fully Bayesian conjugate prior, which allows for analytical posterior calculation and straightforward inference. We derive a variational Bayes inference algorithm and demonstrate the model on the MNIST digits and HGDP-CEPH cell line panel datasets.

• Type: Conference
• Year: 2009
• Where: ICML
• Authors: John William Paisley, Lawrence Carin
• Keywords: Beta Process Factor, Factor Analysis Problem, ICML 2009, Machine Learning, Variational Bayes Inference