Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!helios!tamsun.tamu.edu!jdm5548
From: jdm5548@tamsun.tamu.edu (James Darrell McCauley)
Newsgroups: comp.ai.neural-nets
Subject: finding inputs
Message-ID: <17151@helios.TAMU.EDU>
Date: 11 Jun 91 20:35:33 GMT
Sender: usenet@helios.TAMU.EDU
Followup-To: comp.ai.neural-nets
Organization: Texas A&M University
Lines: 15

A friend ask this question of me and I wasn't quite sure
how to answer. Perhaps someone could help.

  Given a NN, with say, 8 inputs and 3 outputs, that 
  characterizes some process, and if 3 of the inputs
  are functions of other 5, is there a (smart) way to 
  change the inputs to achieve a desired output vector?
  Assume that the net is trained and weights are constant.

Is there a simple solution to this?  How would you
go about this?  What neural network architecture would
work best? Should I read those references on "extracting
rules" posted a while back?

--jdm5548@diamond.tamu.edu, jdm5548@tamagen.bitnet