Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!zombie.ncsc.mil!news.duke.edu!agate!news.mindlink.net!news.bc.net!newsserver.sfu.ca!fornax!kamber
From: kamber@cs.sfu.ca (Micheline Kamber)
Subject: Competitive Learning (rules of thumb?)
Message-ID: <1995Feb25.004446.8215@cs.sfu.ca>
Organization: Faculty of Applied Science, Simon Fraser University
Date: Sat, 25 Feb 1995 00:44:46 GMT
Lines: 18


I'm presently experimenting with the competitive learning
algorithm (by Rumelhart and Zipser, 1985) on some 
data. I was wondering if anyone might be able to
suggest some guidelines for the network's architecture
i.e. if you have used this algorithm,  
  - how may layers did you generally require, and
  - how many clusters per layer, and
  - how may nodes per cluster?

Any rules of thumb (if they exist) or comments on this matter 
would be greatly appreciated.

Many thanks,

Micheline Kamber
kamber@cs.sfu.ca

