Variational Bayesian Independent Component Analysis

Neil D. Lawrence, University of Sheffield
Christopher M. Bishop, Microsoft Research, Cambridge

Abstract

Blind separation of signals through the info-max algorithm may be viewed as maximum likelihood learning in a latent variable model. In this paper we present an alternative approach to maximum likelihood learning in these models, namely Bayesian inference. It has already been shown how Bayesian inference can be applied to determine latent dimensionality in principal component analysis models @Bishop:bayesPCA98. In this paper we derive a similar approach for removing unecessary source dimensions in an independent component analysis model. We present results on a toy data-set and on some artificially mixed images.

  @TechReport{lawrence-ica99, title = {Variational Bayesian Independent Component Analysis}, author = {Neil D. Lawrence and Christopher M. Bishop}, year = {2000}, month = {00}, edit = {https://github.com/lawrennd//publications/edit/gh-pages/_posts/2000-01-01-lawrence-ica99.md}, url = {http://inverseprobability.com/publications/lawrence-ica99.html}, abstract = {Blind separation of signals through the info-max algorithm may be viewed as maximum likelihood learning in a latent variable model. In this paper we present an alternative approach to maximum likelihood learning in these models, namely Bayesian inference. It has already been shown how Bayesian inference can be applied to determine latent dimensionality in principal component analysis models @Bishop:bayesPCA98. In this paper we derive a similar approach for removing unecessary source dimensions in an independent component analysis model. We present results on a toy data-set and on some artificially mixed images.}, key = {Lawrence:ICA99}, linkpsgz = {http://www.thelawrences.net/neil/bica_report.ps.gz}, OPTgroup = {} }
 %T Variational Bayesian Independent Component Analysis %A Neil D. Lawrence and Christopher M. Bishop %B %D %F lawrence-ica99 %P -- %R %U http://inverseprobability.com/publications/lawrence-ica99.html %X Blind separation of signals through the info-max algorithm may be viewed as maximum likelihood learning in a latent variable model. In this paper we present an alternative approach to maximum likelihood learning in these models, namely Bayesian inference. It has already been shown how Bayesian inference can be applied to determine latent dimensionality in principal component analysis models @Bishop:bayesPCA98. In this paper we derive a similar approach for removing unecessary source dimensions in an independent component analysis model. We present results on a toy data-set and on some artificially mixed images. 
 TY - CPAPER TI - Variational Bayesian Independent Component Analysis AU - Neil D. Lawrence AU - Christopher M. Bishop PY - 2000/01/01 DA - 2000/01/01 ID - lawrence-ica99 SP - EP - UR - http://inverseprobability.com/publications/lawrence-ica99.html AB - Blind separation of signals through the info-max algorithm may be viewed as maximum likelihood learning in a latent variable model. In this paper we present an alternative approach to maximum likelihood learning in these models, namely Bayesian inference. It has already been shown how Bayesian inference can be applied to determine latent dimensionality in principal component analysis models @Bishop:bayesPCA98. In this paper we derive a similar approach for removing unecessary source dimensions in an independent component analysis model. We present results on a toy data-set and on some artificially mixed images. ER - 
 Lawrence, N.D. & Bishop, C.M.. (2000). Variational Bayesian Independent Component Analysis.:-