Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornellcs!travelers.mail.cornell.edu!news.tc.cornell.edu!newsserver.sdsc.edu!nic-nac.CSU.net!charnel.ecst.csuchico.edu!psgrain!usenet.eel.ufl.edu!news.mathworks.com!news.kei.com!ddsw1!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: <DAFL7p.Lny@unx.sas.com>
Date: Mon, 19 Jun 1995 17:43:01 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References:  <JFSCHREER.4.2FE59092@BIOLOGY.watstar.uwaterloo.ca>
Organization: SAS Institute Inc.
Keywords: ART2
Lines: 31


In article <JFSCHREER.4.2FE59092@BIOLOGY.watstar.uwaterloo.ca>, JFSCHREER@BIOLOGY.watstar.uwaterloo.ca (Jason F. Schreer) writes:
|> I am using the latest PC version of NeuralWorks Pro II/Plus.  I
|> am interested in grouping a multivariate data set with real
|> numbers (unsupervised learning).  I have used SOM fairly
|> successfully for this task, but I would also like to try some of
|> the ART paradigms. ... What other choices do I have for unsupervised
|> classification of analog data?

The ART algorithms produce degenerate results for noisy data:

   Moore, B. (1988), "ART 1 and Pattern Clustering", in Touretzky, D.,
   Hinton, G. and Sejnowski, T., eds., _Proceedings of the 1988 Connectionist
   Models Summer School_, 174-185,  San Mateo, CA: Morgan Kaufmann.

While this article is about ART1, essentially the same problems occur
with ART2 and FART.

Various clustering algorithms are superior to NN methods for grouping
data. For example:

   Balakrishnan, P.V., Cooper, M.C., Jacob, V.S., and Lewis, P.A. (1994)
   "A study of the classification capabilities of neural networks using
   unsupervised learning: A comparison with k-means clustering",
   Psychometrika, 59, 509-525.

-- 

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.
