Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!yeshua.marcam.com!zip.eecs.umich.edu!umn.edu!milli!young
From: young@milli.cs.umn.edu (Mike Young)
Subject: Re: # of hidden nodes for Radial Basis Function Networks ?
Message-ID: <young.783296572@milli>
Sender: news@news.cis.umn.edu (Usenet News Administration)
Nntp-Posting-Host: milli.cs.umn.edu
Organization: University of Minnesota, Twin Cities
References: <1994Oct23.165843.30929@cc.usu.edu> <Cy7B49.Avw@acsu.buffalo.edu>
Date: Thu, 27 Oct 1994 22:22:52 GMT
Lines: 21

I get the impression that most people are considering the question of # of
nodes within the context of a single layer of RBF units.  Has any work
been done using layers of RBF units?  Given the range of basis 
functions that can be used, there should be some configurations where 
layering would be beneficial.  However, given the undirected nature
of unsupervised learning, you're guaranteed of needing more than those
necessary in a supervised net.  BP suffers from other problems though
(e.g. catastrophic interference), so your choice, as always, depends
on the task being performed.

Mike Young

Dept. of Psychology
317 Elliott Hall
University of Minnesota
Minneapolis, MN 55455
--
Mike Young
Dept. of Psychology
317 Elliott Hall, University of Minnesota
Minneapolis, MN 55455
