Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!news.alpha.net!uwm.edu!math.ohio-state.edu!howland.reston.ans.net!news.sprintlink.net!malgudi.oar.net!utnetw.utoledo.edu!news
From: dweddin@uoft02.utoledo.edu
Subject: RE: # of hidden nodes for Radial Basis Function Networks ?
Message-ID: <CyB9LI.LtK@utnetw.utoledo.edu>
Lines: 48
Sender: news@utnetw.utoledo.edu (News Manager)
Organization: University of Toledo
References: <1994Oct23.165843.30929@cc.usu.edu>
Date: Wed, 26 Oct 1994 17:58:12 GMT

In Article <1994Oct23.165843.30929@cc.usu.edu>
slq15@cc.usu.edu writes:
>	I am working on a character classification problm
>using backprop and radial basis function networks, using Neuralware.
>I am somewhat familiar with backprob, but not much with RBFNs.
>Is there any rule of thumb for finding #of nodes in prototype
>layer in RBFNs ? Seems to me that they use much more hidden nodes
>in RBFN than in BP. I have 35 input features and 62 output classes.
>Typically, I would try 60-80 hidden nodes for BP. Example of RBFN given
>with Neuralware uses 20 nodes in prototype layer for only 2 i/p
>and 3 o/ps, so I was wondering if you need to use many more
>hidden (i.e. prototype layer) nodes for RBFN than in BP. I have
>tried around 200 hidden nodes for RBFN, but performance is much
>worse than BP (90% vs 70%). Are there any other pitfalls or
>fine tuning that should be taken care of for RBFNs ? Any
>comments will be appriciated.
>
>thanks
>Rutvik
>
>--
>Rutvik Desai
>Computer Science
>Utah State University
>rutvik@sys3.cs.usu.edu
> 

I'm doing my research with RBF's and had this problem early on, but then
I switched my method of selecting the centers and radii and drastically
reduced the number of nodes in the hidden layer.

I've found that you use something as simple as the Fuzzy C Means (FCM)
and then choose your radii by finding the distance to the nearest neighbor
and multiplying by a scaling factor between 0 and 1. The number of nodes
needed to get good results has drastically dropped. Has anybody else found
this or am I just being fortunate with my particular data?

Also, I have had some success with the Orthoganal Least Squares algorithm
(OLS) discussed by, Cheng (I think?). Of course, this method requires that
you use a gaussian basis function but since I prefer to use those anyways
this does not affect me.

Donald Wedding
University of Toledo




