Mail Archives: djgpp/1997/03/05/04:40:46
>----------
>From: leathm AT solwarra DOT gbrmpa DOT gov DOT au[SMTP:leathm AT solwarra DOT gbrmpa DOT gov DOT au]
>Sent: 04 March 1997 23:25
>To: fwec AT juno DOT com
>Cc: djgpp AT delorie DOT com
>Subject: Re: neural network code
>
>heh heh, sorry, I was being sarcastic, which maybe I shouldn't have been,
Whats the problem? there aint no thought police around here so there
shouldn't be any
worries
>and no, I dont have a great knowledge of neural networks. I just assumed
>that something the US Army couldn't get running properly would be a little
>hard to code, even it is a simply test base.
Yea this is the problem, Neural nets DO NOT work! they kind of work in a
kind of way, and
they are only designed to model the brains method of cognative function,
not to be the brain.
This is the problem with AI, we cannot model the brain, as we do not
fully understand how it
works! We can only use these models to see if we are on the right track.
> Computers have a simple test ase, either on or off, and I would still say it
>is hard to write something
>like Windows95. In the end everything is simple, just sometimes hard to
>implement... :)
They do use a simple test base true, but they also have powerful
mathematical ability and this is where
the neural net comes in, the test of true or false is made after the
processing. The way it work is difficult
to explain apart from the fact that a nueral network is best categorised
as a statistical device.
Rob Humphris
PS If anyone is interested in the code please mail me direct, as I have
nearly ported it to DJGPP, put the
subject "gimme gimme gimme" so as I can filter the messages out
- Raw text -