Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news4.ner.bbnplanet.net!news.ner.bbnplanet.net!howland.reston.ans.net!vixen.cso.uiuc.edu!uwm.edu!math.ohio-state.edu!magnus.acs.ohio-state.edu!lerc.nasa.gov!kira.cc.uakron.edu!odin.oar.net!malgudi.oar.net!utnetw.utoledo.edu!news
From: dweddin@uoft02.utoledo.edu
Subject: Backprop Question!
Message-ID: <DMJDJ0.90H@utnetw.utoledo.edu>
Lines: 28
Sender: news@utnetw.utoledo.edu (News Manager)
Organization: University of Toledo
Date: Fri, 9 Feb 1996 15:30:56 GMT



Hey all,

have most of my experience in RBF's. Now i'm working with
Backprop. I understand the mechanics, but don't have experience.

I am using a two hidden layer backprop network. can anybody
give me some good rules-of-thumb of how many nodes should be
used for each layer.

should i strive to put most of my nodes in the first hidden layer
and a few in the second?

should most go into the second hidden layer with a few in the
first layer?

should i put about the same number in each of the layers???


any good articles or personal experiences would be much appreciated!

thanks!

Don Wedding, P.E., Ph.D.
Wooden Statue, Inc.


