Re: [computer-go] ANNs as potentially useful in the computergoproblem...

[Don Dailey <drd@xxxxxxx>]

The  book  was  fascinating reading,  but  the  book  was a  lot  more
entertaining than it  was informative.  If it was  about how a machine
learning approach could work as  well or better than a knowledge-based
approach then it fell far short of that.

To  me it  was  simply about  the  adventures of  Blondie  and it  was
exciting reading for that alone.   

It was also  an experiment in human psychology.   The true identify of
Blondie was  hidden, people  were led to  believe Blondie was  a human
female.

The experiment showed that an  evaluation function based on NN's could
improve dramatically, but  it was never shown or  compared against any
other program or approach other than a single game.

What's a little  frustrating is that it's possible  Blondie was better
than  any  program  searching to  the  same  depth,  but no  data  was
published that would even give us a hint.

- Don



   From: Myriam Abramson <mabramso@xxxxxxxxxxxxxxxxx>
   Date: Sun, 11 Jan 2004 13:53:18 -0500

   What Blondie24 was about was not the capability of NNs or genetic
   algorithms but rather how a machine learning approach could work as
   well or maybe better than the knowledge-based approach of Chinook, the
   dominant program in Checkers. 

					    myriam

   _______________________________________________
   computer-go mailing list
   computer-go@xxxxxxxxxxxxxxxxx
   http://computer-go.org/mailman/listinfo/computer-go

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://computer-go.org/mailman/listinfo/computer-go