Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!swrinde!hookup!news.mathworks.com!news.duke.edu!concert!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: Invert, Reverse, ...?
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <Cz66K3.Hpz@unx.sas.com>
Date: Sat, 12 Nov 1994 19:38:26 GMT
References: <Cz0xAD.DK0@unx.sas.com> <3a07io$28g@scapa.cs.ualberta.ca>
Nntp-Posting-Host: hotellng.unx.sas.com
Organization: SAS Institute Inc.
Lines: 40


In article <3a07io$28g@scapa.cs.ualberta.ca>, arms@cs.ualberta.ca (Bill Armstrong) writes:
[various good stuff deleted]
|> In order to invert the results of neural network learning, my suggestion
|> would be to use ALNs, not backprop-type nets.  As a reference to this,
|> you could have a look at atree/wcnnpub.ps.Z on ftp.cs.ualberta.ca .
|>
|> Hope this helps with your macro,

Thanks!

|> though I don't see how you are going
|> to succeed in programming it efficiently without ALNs.

I use a general-purpose nonlinear programming procedure called NLP in
the SAS/OR product for training neural nets. The way this works is that
you write out the equations for computing the network outputs from the
inputs in a programming language resembling PL/1 (or one of my macros
does that for you).  You also write some more statements to compute the
training criterion (sum of squared errors or whatever), and then you
tell NLP to optimize the training criterion with respect to the weights.

To invert (everybody seems to be calling this inversion) the net,
you use the same statements as before for computing the network
outputs. But instead of using the training criterion, you tell NLP
to fix the weights and to constrain each output to the desired value
(these are nonlinear constraints). Then you invent some objective
function and tell NLP to optimize it with respect to the inputs.
This is very simple to do, but my problem is that I have little
information on what objective functions on the inputs would be
useful in real-life applications, hence my original post.

NLP is a very powerful procedure. If anybody wants more information
on it, send email to Wolfgang Hartmann at saswmh@unx.sas.com .

-- 

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.
