On the Effect of the Form of the Posterior Approximation in Variational Learning of ICA Models

Authors:
Alexander Ilin;Harri Valpola
Affiliations:
Neural Networks Research Centre, Helsinki University of Technology, Finland 02015;Laboratory of Computational Engineering, Helsinki University of Technology, Finland 02015
Venue:
Neural Processing Letters
Year:
2005

Citing 10
Cited 5

Ensemble learning for multi-layer networks

NIPS '97 Proceedings of the 1997 conference on Advances in neural information processing systems 10
Independent factor analysis

Neural Computation
Mean-field approaches to independent component analysis

Neural Computation
An unsupervised ensemble learning method for nonlinear dynamic state-space models

Neural Computation
Variational Bayesian learning of ICA with missing data

Neural Computation
Artifact Reduction in Magnetoneurography Based on Time-Delayed Second Order Correlations

Artifact Reduction in Magnetoneurography Based on Time-Delayed Second Order Correlations
Variational learning of clusters of undercomplete nonsymmetric independent components

The Journal of Machine Learning Research
Hierarchical models of variance sources

Signal Processing - Special issue on independent components analysis and beyond
Variational Learning for Switching State-Space Models

Neural Computation
A blind source separation technique using second-order statistics

IEEE Transactions on Signal Processing

Variational and stochastic inference for Bayesian source separation

Digital Signal Processing
Blind separation of nonlinear mixtures by variational Bayesian learning

Digital Signal Processing
Stability and Chaos of a Class of Learning Algorithms for ICA Neural Networks

Neural Processing Letters
Variational Bayesian blind deconvolution using a total variation prior

IEEE Transactions on Image Processing
Practical Approaches to Principal Component Analysis in the Presence of Missing Values

The Journal of Machine Learning Research

Quantified Score

Hi-index	0.00

Visualization

Abstract

We show that the choice of posterior approximation affects the solution found in Bayesian variational learning of linear independent component analysis models. Assuming the sources to be independent a posteriori favours a solution which has orthogonal mixing vectors. Linear mixing models with either temporally correlated sources or non-Gaussian source models are considered but the analysis extends to nonlinear mixtures as well.