RE: ANFIS vs. training FIS with classical optimization methods

From: Ulrich Bodenhofer (ulrich.bodenhofer@scch.at)
Date: Fri Sep 14 2001 - 23:51:46 MET DST

  • Next message: Gerald Saurugg: "Oil Market"

    Dear Barry,

    the same question attracted my interest almost two years
    ago. Since my organization has some people experienced in numerical
    mathematics and optization too, we started a small project dealing
    with that issue. We knew from the beginning that it would not
    be very easy. IMHO, for the following reason:

    1. First-order methods (e.g. gradient descent) could be too slow.
    2. The objective functions often have a lot of local minima/maxima.
    3. A lot of constraints have to be taken into account (at least if
       you require some reasonable properties that help to maintain
       the interpretability of the system).

    So far, the result was a rather promising method which, however,
    still has some limitations. For more details, please visit

       http://www.scch.at/research/projects/AdvLearn/home/

    If you are interested in more information (e.g. detailed comparisons
    between the new method and other established methods), please do
    not hesitate to write to me.

    Best regards,
    Ulrich

    ------------------------------------------------------------------------
    Dr. Ulrich Bodenhofer

    Area Manager Knowledge-Based Technology
    Software Competence Center Hagenberg GmbH
    Hauptstrasse 99 Tel. +43 (0)7236 3343 832
    A-4232 Hagenberg, Austria Fax +43 (0)7236 3343 888

                  http://www.scch.at/people/ulrich.bodenhofer/
                           http://www.bodenhofer.com/
    ------------------------------------------------------------------------

    > -----Original Message-----
    > From: fuzzy-mail@dbai.tuwien.ac.at
    > [mailto:fuzzy-mail@dbai.tuwien.ac.at]On Behalf Of Barry Wiewiora
    > Sent: Tuesday, September 11, 2001 4:29 AM
    > To: Multiple recipients of list
    > Subject: ANFIS vs. training FIS with classical optimization methods
    >
    >
    > Hi all,
    >
    > I was wondering if anybody knows how ANFIS compares to taking a hand
    > designed FIS, and optimizing its parameters using one of the classical
    > optimization methods. The advantage of doing so would be the ability
    > to place as many constraints that match one's huristics on the
    > problem, as well as the ability to design Mamdani systems or system
    > with multiple outputs.
    >
    > Please let me know if you have experienced such approach, and if you
    > have had much success with it. Also, if there are limitations doing
    > that, I would like to know.
    >
    > Thanks,
    > Barry
    >
    > PS: Please CC me at barry_wiewiora@yahoo.com, if possible.
    >
    >
    > ##################################################################
    > ##########
    > This message was posted through the fuzzy mailing list.
    > (1) To subscribe to this mailing list, send a message body of
    > "SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at
    > (2) To unsubscribe from this mailing list, send a message body of
    > "UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL
    > yoursubscription@email.address.com"
    > to listproc@dbai.tuwien.ac.at
    > (3) To reach the human who maintains the list, send mail to
    > fuzzy-owner@dbai.tuwien.ac.at
    > (4) WWW access and other information on Fuzzy Sets and Logic see
    > http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
    > (5) WWW archive:
    http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html

    ############################################################################
    This message was posted through the fuzzy mailing list.
    (1) To subscribe to this mailing list, send a message body of
    "SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at
    (2) To unsubscribe from this mailing list, send a message body of
    "UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
    to listproc@dbai.tuwien.ac.at
    (3) To reach the human who maintains the list, send mail to
    fuzzy-owner@dbai.tuwien.ac.at
    (4) WWW access and other information on Fuzzy Sets and Logic see
    http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
    (5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html



    This archive was generated by hypermail 2b30 : Sat Sep 15 2001 - 00:12:05 MET DST