Variational Bayesian Nonnegative Matrix Factorisation

Download the zip file here.

This folder contains a Matlab implementation of variational Bayes for
KL (Kullback-Leibler) divergence based Non Negative Matrix
Factorisation, as discussed in:


A. T. Cemgil. Bayesian inference in non-negative matrix factorisation models.
Technical Report CUED/F-INFENG/TR.609, University of Cambridge, July 2008.
Accepted for publication to Computational Intelligence and Neuroscience   

Features:
* Full Bayesian treatment of NMF
* Handles Missing data (specified via NaN's)
* Enables model selection via variational lower bound (finding automatically the 'rank' of the decomposition)
* Hyperparameter adaptation (Finding the appropriate sparsity criteria  )
* Compact implementation

To run a demo, simply extract all files to a directory, in Matlab
change to the directory and run the demo:

>>  gnmf_vb_demo01

This generates a positive matrix X from the NMF generative model, then
computes a positive decomposition with hyperparameter adaptation.

Note: This implementation needs the statistics toolbox of
Matlab. Actually, we only use the Poisson and Gamma random variable
generators poissrnd and gamrnd, so you should be able to easily remove
the toolbox dependence.

File List:

gnmf_vb_poisson_mult_fast.m            NMF implementation with bound computation
gnmf_solvebynewton.m                   Hyper parameter optimisation
marginal_plot.m                        Visualisation Utility
parse_optionlist.m                     Parameter parsing utility


%11-May-2009     3:09 PM        ATC  

Author:
Ali Taylan Cemgil
Dept. of Computer Eng., Bogazici University, 
34342 Bebek, Istanbul, Turkey

email <append .edu.tr> :  taylan DOT cemgil AT boun   

%11-May-2009	 3:09 PM	ATC	

Copyright (C) 2009  Ali Taylan Cemgil, 
     
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
     
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.