The followings are additional operators and sets:
- Sugeno Compliment
- Yager Compliment
- Non parameterized T-Norm (AND , OR)
- Non parameterized S-Norm (AND , OR)
- Both Parameterized T-Norm and S-Norm
|
|-- Schweizer & Sklar
|-- Yager
|-- Dubois & Prade
|-- Hamacher
|-- Frank
|-- Sugeno
|-- Dombi- Two sided Pi Fuzzy Set - Gaussian Fuzzy Set - Two sided Gaussian Fuzzy Set (can be non-symmetric) - Generalized Bell Fuzzy Set - Sigmoidal Fuzzy Set - Difference of Two Sigmoidals Fuzzy Set - Product of Two Sigmoidals Fuzzy Set
Currently the implementations used the Java double[] array, but I am going to use JAMA (Java Matrix Algebra) because most of my other stuff (development) is done using that API (JAMA).
http://math.nist.gov/javanumerics/jama/
It will be easier to write the internal implementations in JAMA,
because it is simple doing double floating point calculations
using matrix than Java double[] array (that is less effort of
coding). The other reason is, since the aim is to make the
fuzzy systems adaptive , those fuzzy sets will be almost continuous
that is lots of fuzzy set points (Set Points). It will be easier to
use matrix to calculate the differentiation (first derivative) then doing looping all the time when you are dealing (doing calculations)with the array of set points. Also JOONE (Java Object Oriented Neural Engine) uses a matrix class (not JAMA) for calculations of its input-output weights. I intend to replace that matrix class with the one from JAMA.
Perhaps , I can send you the files of the additonal operators & sets (zipped) when the internal implementations are done in JAMA to be made available from your site for download.
Cheers, Sione.
Orchard, Bob wrote:
Can you identify the additions you made to FuzzyJ ... they might be worth adding for everyone.
Thanks, bob.
Bob Orchard
National Research Council Canada Conseil national de recherches Canada
Institute for Information Technology Institut de technologie de
l'information
1200 Montreal Road, Building M-50 M50, 1200 chemin Montrial
Ottawa, ON, Canada K1A 0R6 Ottawa (Ontario) Canada K1A 0R6
(613) 993-8557 (613) 952-0215 Fax / tilicopieur
[EMAIL PROTECTED] Government of Canada | Gouvernement du Canada
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Behalf Of Sionep
Sent: Thursday, March 25, 2004 4:33 AM
To: [EMAIL PROTECTED]
Subject: JESS: Article :- Adaptive Fuzzy Expert Systems
Hi All,
There is an article that might be interest to some members of this list. The article is not Java or JESS based but it gives good theoretical ground for developing such systems using JESS + FuzzyJ and some extra numerical computing code development effort. The title is :
"Adaptive Fuzzy Expert Systems"
The link is shown below, where the PDF file is found near the bottom of the page for "Publication" section under the "Journal Articles" sub-heading.
http://www.aut.ac.nz/research_showcase/research_activity_areas /kedri/rccidss.shtml#projects
Adaptive Fuzzy Expert Systems has been done before in academic
and industry commercial applications but there are always new techniques emerging all the time. The numerical computing techniques used
was the "gradient descent" that is used in neural network multi-layer
perceptrons.
The way I see state-of-the-art rule-based-systems currently available or
up-and-coming systems will be shifted from static rule-based systems to
ADAPTIVE (self-aware or self-tune) Expert systems. It will be the norm of the near future, and JESS + FuzzyJ will be right at it. This is not my prediction , it is the evolving trend at the moment as seminars on the subject are frequently announced for the "Soft-Computing" mailing list from Berkeley.
I have been interested in this subject (Adaptive or Neuro-Fuzzy expert system) for a while, and I am currently attempting to combine FuzzyJ with JOONE (Java Object Oriented Neural Engine) which is an open source Java Neural Network API. My models works in MatLab but the next step is to try and replicate it in Java using FuzzyJ and JOONE, because some of the functions in MatLab Fuzzy Logic toolbox are not available in FuzzyJ. I have written some extra Java classes to supplement FuzzyJ. Once FuzzyJ and JOONE can work together and then the rest can be done with ease. This means this neuro-fuzzy package will be easily integrated with JESS thus having the ability to develop adaptive fuzzy expert systems in Java.
One of the author's of the article Prof Nik Kasabov has been using CLIPS and FuzzyCLIPS in the past. I do not know whether he is using JESS or not, because he indicated to me last year (2003) that he is keen to get his group (Knowledge Engineering Center) to adopt JESS.
Cheers, Sione.
--------------------------------------------------------------------
To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
in the BODY of a message to [EMAIL PROTECTED], NOT to the list
(use your own address!) List problems? Notify [EMAIL PROTECTED]
--------------------------------------------------------------------
-------------------------------------------------------------------- To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]' in the BODY of a message to [EMAIL PROTECTED], NOT to the list (use your own address!) List problems? Notify [EMAIL PROTECTED] --------------------------------------------------------------------
-------------------------------------------------------------------- To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]' in the BODY of a message to [EMAIL PROTECTED], NOT to the list (use your own address!) List problems? Notify [EMAIL PROTECTED] --------------------------------------------------------------------
