Nonparametric factor analysis with beta process priors

Published

Journal Article

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. Copyright 2009.

Full Text

Duke Authors

Cited Authors

  • Paisley, J; Carin, L

Published Date

  • September 15, 2009

Published In

  • Acm International Conference Proceeding Series

Volume / Issue

  • 382 /

Digital Object Identifier (DOI)

  • 10.1145/1553374.1553474

Citation Source

  • Scopus