Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!hookup!swrinde!pipex!uunet!ulowell.uml.edu!umassd.edu!umassd!s_vogt
From: s_vogt@umassd.cis.umassd.edu (Stefan Vogt)
Subject: Re: Q: ANN & softeware engineering?
Message-ID: <D3Av5L.C6F@umassd.edu>
Sender: usenet@umassd.edu (USENET News System)
Organization: University of Massachusetts Dartmouth
X-Newsreader: TIN [version 1.2 PL2]
References: <D2z423.IKq@umassd.edu> <pq+Zide.predictor@delphi.com>
Date: Wed, 1 Feb 1995 02:34:32 GMT
Lines: 28

Will Dwinnell (predictor@delphi.com) wrote: (thank you for your posting)

:                                           ... There are already
: well understood principles of neural system design which give a
: general prescription for specific situations and problems, if that is
: what you mean. 
That's what I am looking for. Where can I find these principles of
neural system design. I don't want a recipe for a specific problem.
If you have a problem, how to you choose eg. the net architcture, the
learning algorithm,and the data representation? Do you have a certain
decision path in your development? 

: Some of us actually make a living at applying these technologies to
: problems in the "real world", and we do so professionally, effectively
: and economically.
If you implement ANN economically, you need a plan or method when to
make a decision, when to stop eg. the training... Could you send me
information about this? 

: The fact is that applying neural networks, or any other software
: technology (relational database technology, simulation, etc.) requires
: a certain degree of planning and supervision.  This is exactly what we
: and many others do every day.
The development of ANNs is sometimes a little more uncertain than other
software, isn't it?

Bye, Stefan

