Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!godot.cc.duq.edu!newsgate.duke.edu!news.mathworks.com!newsfeed.internetmci.com!zdc!news1.noc.netcom.net!noc.netcom.net!slave.strata.com!synapse!dxn
From: dxn@strata.com (Deepak Nulu)
Subject: Re: Fuzzy RULES in FUZZY ART???
Message-ID: <DsHK3u.Btt@strata.com>
Sender: news@strata.com
Reply-To: dxn@strata.com
Organization: StrataCom, Inc
References: <4ok8f7$c0q@brachio.zrz.TU-Berlin.DE>
Date: Tue, 4 Jun 1996 17:19:05 GMT
Lines: 42

In article c0q@brachio.zrz.TU-Berlin.DE, Hadi Widjaja <hadicagi@sp.zrz.tu-berlin.de> () writes:
>I'm trying to understand the architecture of Fuzzy Art but
>in this far I don't see any implementation of fuzzy rules
>in this network. Suppose they are represented by the weights?
>My questions are:
>
>1. How to realize (implement) fuzzy rules in to Fuzzy Art?
>2. How to interprete fuzzy reasoning mechanism?
>
>Thank you in advance.
>
>
>
>
>Hadi
>

As far as I understand it, the only reason it is called Fuzzy ART is
due to the use of the fuzzy equivalents of the bitwise OR and bitwise
AND operators, namely, MAX and MIN. There is really nothing related to
fuzzy logic as such.

Fuzzy ART functions as a clustering algorithm, with the cluster boundaries
being defined as hyper-rectangles. The boundaries are crisp. The weights
represent the opposite corners of the rectangles.

There was a similar implementation by someone (not from Boston U) wherein
the boundaries were fuzzy. Send me email if you are really interested and
I will dig up the reference.

- deepak

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Deepak Nulu			The views expressed above, if any :-), are
				mine & not necessarily those of StrataCom.
StrataCom, Inc			
1400 Parkmoor Avenue		dnulu@strata.com
San Jose, CA 95126		(408) 882-2043

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

