readme

Nonlinear dynamical factor analysis Matlab package

Nonlinear Dynamic Factor Analysis Matlab package
version 1.0, 2005-08-12

This package is based on the Nonlinear Factor Analysis Matlab package

WHAT'S NEW------------------------------------------------------------

- New approximation for the tanh nonlinearity based on Gauss-Hermite 
  quadratures which should solve most of the stability problems

- Improved learning algorithm based on Natural Conjugate Gradient

- Mex files are no longer used, recent versions of Matlab should run 
  the code fast enough as it is

- Support for missing values

- Improved initialization

- Numerous code cleanups

USAGE-----------------------------------------------------------------

[sources, net, tnet, params, status] = ...
	NDFA(data, 'searchsources', 5, 'hidneurons', 30, 'thidneurons', 20);

Extract 5 nonlinear factors from data using an
observation MLP with 30 hidden neurons and temporal MLP
with 20 hidden neurons and default source initialisation.
  

result = NDFA(data, 'initsources', my_s, 'hidneurons', 30, 'thidneurons', 20, 'iters', 50);

Extract nonlinear factors from data using an observation
MLP with 30 hidden neurons and temporal MLP with 20
hidden neurons and custom initialisation given by my_s
using 50 iterations of the algorithm.


result = NDFA(data, result, 'iters', 500);

Continue the previous simulation for 500 more iterations.


result = NDFA(data, 'searchsources', 6, 'initcontrol', my_u, 'hidneurons', 30, 'thidneurons', 20);

Extract 6 nonlinear factors from data driven by control 
my_u using an observation MLP with 30 hidden neurons and
temporal MLP with 20 hidden neurons.


result = NDFA(data, 'searchsources', 4, 'hidneurons', 30, 'thidneurons', 20, 'notimedep', 500);

Extract 4 nonlinear factors from two part data with the 
first part containing 500 samples using an observation 
MLP with 30 hidden neurons and a temporal MLP with 20 
hidden neurons.


result = ...
    NDFA(data, 'searchsources', 5, 'hidneurons', 30, 'thidneurons', 20);

Extract 5 nonlinear factors from data using an
observation MLP with 30 hidden neurons and temporal MLP
with 20 hidden neurons and zero source initialisation.

LEGALESE--------------------------------------------------------------

Copyright (C) 2002-2005 by Harri Valpola, Antti Honkela and Matti Tornio

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.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.


CONTACT INFORMATION---------------------------------------------------

The WWW home page of this package is:

  http://www.cis.hut.fi/projects/ica/bayes/

The authors can be reached by paper mail:

  Bayes group
  Laboratory of Computer and Information Science
  P.O.Box 5400
  FI-02015 HUT
  Finland

or by email:

  [e-mail]