"Ralf Callenberg" wrote in message
...
14.04.2007 21:24, Anders Thulin:

But isn't the principle the same as Samuels used in his checkers
program?

I didn't find anything detailed about Samuel's program. In the texts I

Original 1959 paper:

http://domino.research.ibm.com/tchjr...4?OpenDocument

Followup 1967 paper:

http://domino.research.ibm.com/tchjr...6?OpenDocument

2000 reprint:

http://domino.research.ibm.com/tchjr...5?OpenDocument


found is just mentioned, that his evaluation function is some ploynomial.
But what are the constituents of this function is nowhere given. Were they
very general or was already some knowledge about what is important in
checkers introduced?

Greetings,
Ralf




----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----

On Apr 16, 3:03 pm, "Hello" wrote:


Original 1959 paper:

http://domino.research.ibm.com/tchjr...b9bc46ed06cbac...

Followup 1967 paper:

http://domino.research.ibm.com/tchjr...b9bc46ed06cbac...

2000 reprint:

http://domino.research.ibm.com/tchjr...b9bc46ed06cbac...

Thanks a lot! It's a nice read.

Now some questions from former postings can be answered: the terms in
the polynomial in the evaluation function are indeed checkers
specific. Samuel's program then adapts the coefficiants in a learning
process. This is realized by letting play engines with different
coefficients against each other and also by comparing the found moves
by the program with moves from master games. There is an interesting
remark in the introduction to the first paper: a "Neural Network
approach [...] should lead to the development of general-purpose
learning machines". But as this appearently was beyond existing
technology ("we have a long way to go before we obtain practical
devices"), he uses the more specific representation.

Greetings,
Ralf