%% Version 0.0: Feb 5th 2009

This package contains MATLAB code for fitting models of marginal
independence for continuous data using a 'factorial mixture of
Gaussians'. For more detais about the model see

* Silva, R and Ghahramani, Z. (2009). ``Factorial mixture of Gaussians and the 
  marginal independence model''. Proceedings of the Twelfth International Conference on 
  Artificial Intelligence and Statistics, AISTATS.

This code requires MATLAB's optimization package.

File 'run.m' contains the main routine. See the file for documentation
concerning the inputs and outputs.

You might need to add all subdirectories in 'src' to your path. Change directory
to the one storing 'run.m' and enter the following:

path(path,'graphlib');
path(path,'lvm');
path(path,'relax');
path(path,'other');

You will need to provide a bi-directed graph (in matrix format, as
detailed in 'run.m'). One simple way of learning structure is
provided by the function 'graph_learning' (you will need to
download other package. See the file 'graph_learning.m')

Comments, suggestions and bug reports:

Ricardo Silva
ricardo@stats.ucl.ac.uk

----
London, February 5th 2009
