Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!hookup!caen!msunews!harbinger.cc.monash.edu.au!news.cs.su.oz.au!metro!unsw.edu.au!solution!jim
From: jim@solution ()
Subject: NN for cluster analysis
Message-ID: <1995Jan2.043117.23644@usage.csd.unsw.OZ.AU>
Sender: news@usage.csd.unsw.OZ.AU
Nntp-Posting-Host: solution.maths.unsw.edu.au
Organization: University of New South Wales, Australia
Date: Mon, 2 Jan 1995 04:31:17 GMT
Lines: 11

Suppose I take an autoassociative  N-K-N  net (let's say with K
much smaller than N); with the K hidden neurons being radial
basis function neurons, and the N output neurons being linear
(and with the usual backprop training).

Does this give me K-means clustering (of the input patterns),
with the cluster centers being the fan-in vectors of the
hidden neurons?

James Franklin
University of New South Wales
