NFIDENT
              A program for neuro-fuzzy function approximation

                   Copyright (C) 1997-2005 Detlef Nauck 


   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 1, 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., 675 Mass Ave, Cambridge, MA 02139, USA
   or go to http://www.gnu.org/licenses/gpl.html.


NFIDENT implements the NEFPROX approach and to some extent the ANFIS model
to approximate functions by fuzzy systems.

There is no documentation, yet. The NEFPROX algorithm is documented
in http://fuzzy.cs.uni-magdeburg.de/~nauck/nauck.pdf
Detlef Nauck. Data Analysis with Neuro-Fuzzy Methods. 
Habilitation Thesis, University of Magdeburg, 2000.

For other (older) papers on NEFPROX please refer to our web site
at http://fuzzy.cs.uni-magdeburg.de and download one of our papers. E.g.:

@InProceedings{nauck97c,
  author =       {Nauck, Detlef and Kruse, Rudolf},
  title =        {Function Approximation with NEFPROX},
  pages =        "316--323",
  booktitle =    "Proc.\ 4th GI--Workshop on Fuzzy--Neuro Systems",
  year =         {1997},
  publisher =    "infix",
  address =      "Soest"
}
URL: ftp://fuzzy.cs.uni-magdeburg.de/pub/papers/fns97.ps.gz


For ANFIS please refer to J.-S. Roger Jang's publications. E.g: 
@ARTICLE{jang93,
        AUTHOR             = {Jang, J.-S. Roger},
        JOURNAL            = {IEEE Trans.\ Systems, Man \& Cybernetics},
        PAGES              = {665--685},
        TITLE              = {{ANFIS}: Adaptive-Network-Based Fuzzy 
			      Inference Systems},
        VOLUME             = {23},
        YEAR               = {1993}
}

His complete implementation of ANFIS can be obtained from ftp.cs.cmu.edu in
/user/ai/areas/fuzzy/systems/anfis.
NFIDENT currently just implements backpropagation for ANFIS.

How to use NFIDENT:

1) Under Unix: Compile with make (you probably must edit the Makefile)
   Under Windows: NFIDENT runs as console (MS-DOS) application.
   You can compile, for example, with Microsoft Visual C++. Create
   a new Win32 console project and import all .c and .h files.
   See also NOTES under 6).

2) Usage: 
       Learning: nfident -l <command-file>
       Testing : nfident -t <net-file> <data-file> [<result-file>]
       Create gnuplot file from network: nfident -g <net-file> [<gnu-file>]

   Run "nefprox -l mg.ini" to observe how NFIDENT learns the Mackey-Glass
   function. Observe the resulting fuzzy sets with "gnuplot mg.gnu" 
   and the performance with "gnuplot mgplotres.gnu". The latter one only
   works if you use the filenames as set in the mg.ini file. Otherwise
   edit mgplotres.gnu to suit your setup.
   Gnuplot can be obtained from http://www.gnuplot.info/

   Type just "nfident" to obtain a note on how to invoke NFIDENT.
   You can stop the program during learning with CTRL-C without using
   your results.
   An executable is only provided for Window systems - nfident.exe is
   a WIN32 console application and runs in a command window.

3) Study mg.ini and mg*.dat to understand how to use NFIDENT and how
   to influence the learning process.

4) Don't build your hopes on updates. This one took five years ;-)

5) The previous version was 0.91 beta and had one bug fixed concerning
   computation of center of gravity defuzzification.
   (thanks to Harald Zainzinger for pointing out this bug)
   The current version is 0.95 beta and now uses the performance
   measure and rule learning algorithm described in 
   http://fuzzy.cs.uni-magdeburg.de/~nauck/nauck.pdf pages 58 and 60.
   (thanks to Thomas Bauer for letting me know that the available
    version 0.91 beta had a wrong mamdani.c file that was still 
    using the old algorithm from 1997).

6) NOTE
	- Trapezoidal fuzzy sets are not actually implemented
	  in NFIDENT.
	- in fuzzy.c the function isnan is used. For Unix
	  systems that has to be isnan and for Visual C++
	  it has be written as _isnan.

7) Please send bug reports / suggestions / improvements to nauck@ieee.org.

Have fun!

Detlef Nauck
-----------------------------------------------
E-Mail: nauck@ieee.org
WWW   : http://fuzzy.cs.uni-magdeburg.de/~nauck