Path: utzoo!attcan!uunet!snorkelwacker!think!zaphod.mps.ohio-state.edu!uakari.primate.wisc.edu!ames!pasteur!ahmad@icsib8
From: ahmad@icsib8 (Subutai Ahmad)
Newsgroups: comp.ai.neural-nets
Subject: Re: NN and games
Message-ID: <23125@pasteur.Berkeley.EDU>
Date: 16 Mar 90 19:34:32 GMT
References: <90074.094831LAZBM@CUNYVM.BITNET>
Sender: news@pasteur.Berkeley.EDU
Reply-To: ahmad@icsib8 (Subutai Ahmad)
Organization: International Computer Science Institute
Lines: 16
In-reply-to: LAZBM@CUNYVM (laszlo grunfeld)

In article <90074.094831LAZBM@CUNYVM.BITNET>, LAZBM@CUNYVM (laszlo grunfeld) writes:
>
>  Have neural nets been trained to play games or learned evaluation functions?
>Any papers written on the subject?
>

Check out the article "A parallel network that learns to play
backgammon", by G. Tesauro and T. Sejnowski. Aritifical Intelligence
Vol 39, #3 1989.

The network went on to win the backgammon competition at a computer
games tournament in London.  For a short description of this see
"Neurogammon wins Computer Olympiad", by G. Tesauro, Neural
Computation, Vol 1, #3, 1989.

--Subutai Ahmad