Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!news.bluesky.net!news.sprintlink.net!redstone.interpath.net!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: ART2?
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <DAHr5F.Bzn@unx.sas.com>
Date: Tue, 20 Jun 1995 21:46:27 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References: <JFSCHREER.4.2FE59092@BIOLOGY.watstar.uwaterloo.ca> <DAFL7p.Lny@unx.sas.com> <DAHA0L.9Gn@news.tcd.ie> <3s7b06$3nv@news.bu.edu>
Organization: SAS Institute Inc.
Lines: 23


In article <3s7b06$3nv@news.bu.edu>, laliden@cochlea.bu.edu (Lars Liden) writes:
|> Frank Kelly (kellyfj@tcd.ie) wrote:
|> : In <DAFL7p.Lny@unx.sas.com> saswss@hotellng.unx.sas.com (Warren Sarle) writes:
|>
|> : >The ART algorithms produce degenerate results for noisy data:
|>
|> The ART1 and ART2 category proliferation problem discovered by Moore
|> (1988) is valid, however, as Frank suggested, these problems were
|> subsequently addressed (and readily solved) through the use of
|> compliment coding of the input vectors.  ART1, ART2 and FuzzyART
|> remain completely stable under noisy conditions (do not suffer from
|> the category proliferation problem) when compliment coding is used.

As long as you use maxima or minima instead of a consistent estimate
of central tendency for describing clusters, the results will
degenerate with noisy data. Compliment coding is irrelevant.

-- 

Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.
