Re: Can neuro-fuzzy improve the robustness of NN?

david_olmsted@my-dejanews.com
Thu, 4 Feb 1999 20:32:00 +0100 (MET)

In article <36B80F4B.8316218A@cityu.edu.hk>,
meyhpeng@cityu.edu.hk wrote:
> Dear Colleagues,
>
> It is well known that nero-fuzzy system is a attemp to extract fuzzy
> rules from numerical data. But can it improve the robustness or
> performances of neural network by adding membership function layer?
>
> I think it should be very important problem that how to improve the
> robutness of neural network for application under uncertainty. I am
> doing some works on this areas currently.
>
> Maybe it is a way to achieve thi objective by adding fuzzy membership
> functions into neural network, but no paper to prove it, as I know. If
> you have some information regarding this filed, pls let me know. Thanks!
>
> --
> ------------------------------------------------------
> Dr. Y. H. PENG
> MEEM Dept., City Univ. of HongKong
> Email: meyhpeng@cityu.edu.hk
> -------------------------*-----------------------------------
>
>
Dr. Peng,

Neural networks can certainly be improved and made more robust by using fuzzy
logic operations in place of threshold operations in neural networks,
especially since these operations can be constructed out of simpler addition
and subtraction operations. To be specific I have designed such a network
which implements the unsupervised learning strategy of Stephen Grossberg's
ART network but does so more robustly. Since the certainty values are
conserved (meaning that they pass through the network without corruption)they
can be fed into a second layer supervised learning network. This two layer
network can thus learn to seperate patterns into any arbitrary class.

Since I am proposing that this network represents the tectum (first layer)and
reticular formation (second layer)in vertebrates it can be found in the
tectum-retecular formation page at http://www.neurocomputing.org.

Sincerely,
David Olmsted

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own

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