Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!pipex!uknet!info!news
From: your.cs9h2hjm@Swansea.ac.uk (John Hammond)
Subject: Help: Layers in BackProp neural nets
Message-ID: <Cz9ICn.2Ep@info.swan.ac.uk>
Sender: news@info.swan.ac.uk
Nntp-Posting-Host: cc14612.swan.ac.uk
Reply-To: cs9h2hjm@Swansea.ac.uk
Organization: Swansea University
X-Newsreader: WinVN 0.90.5
Date: Mon, 14 Nov 1994 14:45:58 GMT
Lines: 19

I'm currently writing a piece of software to aid in the 
construction of neural networks - it's my final year project.

I'm mainly dealing with back-propagation.

Can anybody tell me if there is any general rule of thumb 
for deciding on the topology of the networks required to 
carry out a given task. I especially want to know how to 
decide on the number of layers a network, needs.  Is there
any way of determining this apart from trial and error??

Reply via e-mail please.

Thanks 

John Hammond
cs9h2hjm
Why should I put a sig, nobody cares about it anyway!

