Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornellcs!newsstand.cit.cornell.edu!newstand.syr.edu!news.maxwell.syr.edu!mr.net!news.sgi.com!news.sprintlink.net!news-peer.sprintlink.net!news.sprintlink.net!news-hub.sprintlink.net!news.sprintlink.net!news-ana-24.sprintlink.net!interpath!news.interpath.net!news.interpath.net!sas!newshost.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: Neural Net Handbook
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <E1z6CG.ID2@unx.sas.com>
Date: Fri, 6 Dec 1996 04:41:04 GMT
X-Nntp-Posting-Host: hotellng.unx.sas.com
References:  <584vtr$mjg@larch.cc.swarthmore.edu>
Organization: SAS Institute Inc.
Lines: 33


In article <584vtr$mjg@larch.cc.swarthmore.edu>, haneef@engin.swarthmore.edu (Omar Haneef '96) writes:
|> 
|> I am a recent entrant intot the ranks of connectionist modellers, so perhaps
|> I am missing something, but is there (there certainly OUGHT to be) a source
|> book that contains principles for achieving Neural Net outputs? I mean some
|> mammoth source that would have general principles such as "IF you have back
|> prop AND you have three layers THEN to decrease activation train on 0", 
|> except it would be more thorough and more systematic? This way anybody who
|> wants a particular output or performance does not have to guess their way
|> to it.

No, there isn't any such handbook. If there were, the FAQ could be
a lot shorter!

And, unfortunately, there is not going to be any such handbook in the
foreseeable future. The problem is that the best way to do almost
anything regarding NNs depends on the particular data or problem that
you are trying to solve. And if you knew enough about the data to know
the best way to tackle it with a NN, then you probably wouldn't need a
NN in the first place. For a very extensive demonstration of "it
depends on the data", see:

   Michie, D., Spiegelhalter, D.J. and Taylor, C.C. (1994), Machine
   Learning, Neural and Statistical Classification, Ellis Horwood. 

-- 

Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.
 *** Do not send me unsolicited commercial or political email! ***

