RE: ANFIS training

From: Li Xiang Ling (xli@sckcen.be)
Date: Fri Sep 28 2001 - 15:21:35 MET DST

  • Next message: michal: "Re: fuzzy and assembly"

    This message is in MIME format. Since your mail reader does not understand
    this format, some or all of this message may not be legible.

    --------------InterScan_NT_MIME_Boundary
    Content-Type: multipart/alternative;
            boundary="----_=_NextPart_001_01C145C9.72E3A870"

    ------_=_NextPart_001_01C145C9.72E3A870
    Content-Type: text/plain;
            charset="iso-8859-1"

    -----Original Message-----
    From: Tom De Muer [mailto:tom.demuer@skynet.be]
    Sent: dinsdag 25 september 2001 15:31
    To: Multiple recipients of list
    Subject: ANFIS training

    Hello,

    I've read the article of R. Jang about the ANFIS architecture. If I'm
    correct the network is (or can be) trained with a gradient descent. How
    does this work: I think it has to be something like this:
    - make f(input,a,b,c,...)
    - derive df/da (partial derivatives)
    - the adaptation of a is delta_a = etha * (target - f(input,a,b,c...))

    Is there another way : I was thinking of building some elementary building
    blocks like a multiplier, a summation, ... which are connected and then get
    trained by backpropagation. This works fine with relative simple functions
    but my ANFIS network diverges from the solution. Is it possible to connect
    random building blocks (pure tree, no loops) and train them by
    backpropagation or are there limitations?

    Thanks,
    Tom De Muer

    ############################################################################
    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

    ------_=_NextPart_001_01C145C9.72E3A870
    Content-Type: text/html;
            charset="iso-8859-1"
    Content-Transfer-Encoding: quoted-printable

    <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
    <HTML>
    <HEAD>
    <META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
    charset=3Diso-8859-1">
    <META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
    5.5.2653.12">
    <TITLE>RE: ANFIS training</TITLE>
    </HEAD>
    <BODY>
    <BR>
    <BR>

    <P><FONT SIZE=3D2>-----Original Message-----</FONT>
    <BR><FONT SIZE=3D2>From: Tom De Muer [<A =
    HREF=3D"mailto:tom.demuer@skynet.be">mailto:tom.demuer@skynet.be</A>]</F=
    ONT>
    <BR><FONT SIZE=3D2>Sent: dinsdag 25 september 2001 15:31</FONT>
    <BR><FONT SIZE=3D2>To: Multiple recipients of list</FONT>
    <BR><FONT SIZE=3D2>Subject: ANFIS training</FONT>
    </P>
    <BR>

    <P><FONT SIZE=3D2>Hello,</FONT>
    </P>

    <P><FONT SIZE=3D2>I've read the article of R. Jang about the ANFIS =
    architecture.&nbsp; If I'm</FONT>
    <BR><FONT SIZE=3D2>correct the network is (or can be) trained with a =
    gradient descent.&nbsp; How</FONT>
    <BR><FONT SIZE=3D2>does this work: I think it has to be something like =
    this:</FONT>
    <BR><FONT SIZE=3D2>- make f(input,a,b,c,...)</FONT>
    <BR><FONT SIZE=3D2>- derive df/da (partial derivatives)</FONT>
    <BR><FONT SIZE=3D2>- the adaptation of a is delta_a =3D etha * (target =
    - f(input,a,b,c...))</FONT>
    </P>

    <P><FONT SIZE=3D2>Is there another way : I was thinking of building =
    some elementary building</FONT>
    <BR><FONT SIZE=3D2>blocks like a multiplier, a summation, ... which are =
    connected and then get</FONT>
    <BR><FONT SIZE=3D2>trained by backpropagation.&nbsp; This works fine =
    with relative simple functions</FONT>
    <BR><FONT SIZE=3D2>but my ANFIS network diverges from the =
    solution.&nbsp; Is it possible to connect</FONT>
    <BR><FONT SIZE=3D2>random building blocks (pure tree, no loops) and =
    train them by</FONT>
    <BR><FONT SIZE=3D2>backpropagation or are there limitations?</FONT>
    </P>

    <P><FONT SIZE=3D2>Thanks,</FONT>
    <BR><FONT SIZE=3D2>Tom De Muer</FONT>
    </P>
    <BR>
    <BR>
    <BR>
    <BR>

    <P><FONT =
    SIZE=3D2>###############################################################=
    #############</FONT>
    <BR><FONT SIZE=3D2>This message was posted through the fuzzy mailing =
    list.</FONT>
    <BR><FONT SIZE=3D2>(1)&nbsp; To subscribe to this mailing list, send a =
    message body of</FONT>
    <BR><FONT SIZE=3D2>&quot;SUB FUZZY-MAIL myFirstName mySurname&quot; to =
    listproc@dbai.tuwien.ac.at</FONT>
    <BR><FONT SIZE=3D2>(2)&nbsp; To unsubscribe from this mailing list, =
    send a message body of</FONT>
    <BR><FONT SIZE=3D2>&quot;UNSUB FUZZY-MAIL&quot; or &quot;UNSUB =
    FUZZY-MAIL yoursubscription@email.address.com&quot; </FONT>
    <BR><FONT SIZE=3D2>to listproc@dbai.tuwien.ac.at</FONT>
    <BR><FONT SIZE=3D2>(3)&nbsp; To reach the human who maintains the list, =
    send mail to</FONT>
    <BR><FONT SIZE=3D2>fuzzy-owner@dbai.tuwien.ac.at</FONT>
    <BR><FONT SIZE=3D2>(4)&nbsp; WWW access and other information on Fuzzy =
    Sets and Logic see</FONT>
    <BR><FONT SIZE=3D2><A =
    HREF=3D"http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info" =
    TARGET=3D"_blank">http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.in=
    fo</A></FONT>
    <BR><FONT SIZE=3D2>(5) WWW archive: <A =
    HREF=3D"http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html" =
    TARGET=3D"_blank">http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/inde=
    x.html</A></FONT>
    </P>

    </BODY>
    </HTML>
    ------_=_NextPart_001_01C145C9.72E3A870--

    --------------InterScan_NT_MIME_Boundary--

    ############################################################################
    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 : Fri Sep 28 2001 - 15:39:38 MET DST