readme.txt

这是核函数独立分量分析KICA的源码和相关程序

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


Version 1.2 - July 7th, 2003
------------------------------






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, Journal of Machine Learning Research, 3, 1-48, 2002.

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 [e-mail]. The
current version 1.2 has been released on July, 7th 2003. It has been
tested on both matlab 5 and matlab 6.  Check regularly the following for
newer versions: http://www.cs.berkeley.edu/~fbach


The package also includes functions to sample from the distributions used
in the JMLR paper (folder 'distributions').



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

and

 >> mex chol_hermite.c

It should create compiled files whose extensions depends on the platform
you are using:
      Windows: chol_gauss.dll     and  chol_hermite.dll 
      Solaris: chol_gauss.mexsol  and  chol_hermite.dll
      Linux  : chol_gauss.mexglx  and  chol_hermite.dll

To check if the file was correcly compiled, type

 >> which chol_gauss
 >> which chol_hermite

and the name of the compiled versions 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 functions that you should use to run the ICA algorithm are
'kernel_ica' (a function with a default setting of parameters)
and 'kernel_ica_options' (where various options can be tried).
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 with Gaussian
kernel in one or higher dimensions

 -chol_hermite.c/.m : incomplete cholesky decomposition with Hermite
polynomial kernel in (currently) only one dimension





Package file list
-----------------

amari_distance.m    : Amari distance between two square matrices
chol_gauss.c        : incomplete cholesky (Gaussian kernel) - C source
chol_gauss.m        : incomplete cholesky (Gaussian kernel) - M file
chol_hermite.c      : incomplete cholesky (Hermite kernel) - C source
chol_hermite.m      : incomplete cholesky (Hermite kernel) - M file
chol_poly.c         : incomplete cholesky (Polynomial kernel) - C source
chol_poly.m         : incomplete cholesky (Polynomial kernel) - M file
contrast_emp_grad.m : derivative of m-way contrast functions
contrast_emp_grad_oneunit.m : derivative of the one-unit contrast functions
contrast_ica.m      : m-way contrast functions
contrast_update_oneunit.m : one-unit contrast function
demo_kernel_ica.m   : demonstration script
empder_search.m     : local search (reaches a local minimum) - full contrast
empder_search_oneunit.m : local search (reaches a local minimum) - one-unit contrast
global_mini.m       : global minimization with random restarts - full contrast
global_mini_oneunit.m : global minimization with random restarts - one-unit contrast
global_mini_sequential.m : global minimization with random restarts
                           one-unit contrast + deflation scheme
kernel_ica.m        : performs ICA using the kernel ICA algorithm with no options
kernel_ica_options.m: performs ICA using the kernel ICA algorithm with options

rand_orth.m         : generates random matrix with orthogonal columns
update_contrast.m   : used for efficient computation of empirical gradient

distributions/usr_distrib.m : function to sample from 18 predefined distributions.