Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!europa.chnt.gtegsc.com!gatech!howland.reston.ans.net!newsjunkie.ans.net!butch!enterprise!news
From: Don Specht <specht@pc-smtp.rdd.lmsc.lockheed.com>
Subject: Re: PNN vs Classification Trees (vs what...?)
Content-Type: text/plain; charset=us-ascii
Message-ID: <1995Aug3.211556.26876@enterprise.rdd.lmsc.lockheed.com>
Sender: news@enterprise.rdd.lmsc.lockheed.com (News Administrator)
Content-Transfer-Encoding: 7bit
Organization: Lockheed Martin Research Labs
X-URL: news:ok7y6IG00h4vMS7l9n@andrew.cmu.edu
References: <ok7y6IG00h4vMS7l9n@andrew.cmu.edu>
Date: Thu, 3 Aug 95 21:15:56 GMT
MIME-Version: 1.0
X-Mailer: Mozilla 1.1N (Macintosh; I; 68K)
Lines: 9

When using PNN, you will of course need to do some type of clustering and 
use the cluster centers (weighted by number of patterns represented) as 
nodes in the PNN.   
The approach which yields the most-compact set of overlapping clusters is 
probably the one described in "Maximum Likelihood Training of Probabilistic 
Neural Networks," by Streit and Luginbuhl, IEEE Trans. on Neural Networks, 
Vol 5, pp. 764-783.


