+------------+
| Kernel ICA |
+------------+


Version 1.0 - March 19th, 2002
------------------------------






Description
-----------

The kernel-ica package is a  Matlab program that implements the Kernel
ICA algorithm for independent component analysis (ICA). The Kernel ICA
algorithm is based on the minimization of a contrast function based on
kernel ideas. A contrast function measures the statistical dependence
between components, thus when applied to estimated components and
minimized over possible demixing matrices, components that are as
independent as possible are found. For more information, please read the
following paper:

Francis R. Bach, Michael I. Jordan (2001). Kernel Independent Component
Analysis, Technical Report UCB//CSD-01-1166, University of California,
Berkeley.

The kernel-ica package is Copyright (c) 2002 by Francis Bach. If you
have any questions or comments regarding this package, or if you want to
report any bugs, please send me an e-mail to fbach@cs.berkeley.edu. The
current version 1.0 has been released on March, 19th 2002. It has been
tested on both matlab 5 and matlab 6.  Check regularly the following for
newer versions: http://www.cs.berkeley.edu/~fbach





Installation
------------

1. Unzip all the .m files in the same directory



2. (Optional) if you want a faster implementation which uses pieces of C
code: at the matlab prompt, in the directory where the package is
installed, type:

 mex chol_gauss.c

It should create a compiled file whose extension depends on the platform
you are using:
 Windows: chol_gauss.dll
 Solaris: chol_gauss.mexsol
 Linux  : chol_gauss.mexglx
To check if the file was correcly compiled, type

 which chol_gauss

and the name of the compiled version should appear. If you have any
problems with the C file of if you are using a platform i did not
mention, please e-mail me.





How to use the kernel-ica package
---------------------------------

The function that you should use to run the ICA alforithm is
'kernel_ica'. a detailed description of its options are described inside
the file and can be reached by simply typing 'help kernel_ica' at the
matlab prompt. A simple demonstration script is provided :
'demo_kernel_ica'.

NB: all the data should be given in columns, that is, if you have m
components and N samples, the matrix should be m x N.


If you wish to investigate the tools and methods we used for this
algorithms, you will find the following files useful:

 -contrast_ica.m  : computation of the contrast functions based on
Kernel canonical correlations

 -chol_gauss.c/.m : incomplete cholesky decomposition