Xref: utzoo rec.games.misc:15287 comp.ai:8909
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!batcomputer!cornell!uw-beaver!mit-eddie!media-lab!minsky
From: minsky@media-lab.MEDIA.MIT.EDU (Marvin Minsky)
Newsgroups: rec.games.misc,comp.ai
Subject: Re: How do games learn? (was  Re: looking for name of a game)
Message-ID: <5590@media-lab.MEDIA.MIT.EDU>
Date: 2 Apr 91 23:23:09 GMT
References: <91090.130516ASNXS@ASUACAD.BITNET> <1991Apr1.130740.19670@IDA.ORG> <16087.27f874ec@levels.sait.edu.au>
Reply-To: minsky@media-lab.media.mit.edu (Marvin Minsky)
Organization: MIT Media Lab, Cambridge MA
Lines: 13

In article <16087.27f874ec@levels.sait.edu.au>
marwk@levels.sait.edu.au writes in regard to Samuel's Checker-learning method
>
>How can one write a program that learns?  I do not see how.
>One has parameters in the heuristic evaluation functions, so how can these
>change based on experience?

This is explained in the classic paper, A. L. Samuel, "Some studies in
machine learning using the game of checkers," IBM J. Res. 
Dev.,vol. 3,pp. 211-219;July,1959.  There is a summary of this in my
"Steps toward AI" paper which you can find reprinted in a book called
Computers and Thought, by Feigenbaum and Feldman.  Samuel uses several
different learning methods,  and has an important discussion of the
"credit assignment" problem.