It is true that MC-programs has a bias towards overconcentration. But...
1) A improvements to the simulations of MC-program as implemented by MoGo and
Valkyria does diminish the problem. In fact most of the strength of these
programs from doing that. I think it is next to possible to explicitly program
the concept of overconcentration, but fortunately you do not have to. The
problem goes away when you fixed other more immediate problems with the
simulations.
2) Deeper search makes these programs stronger. Valkyria always get stronger
with increased search. Just playing MoGo on fixed time limits does not
tell you
anything about what it could possibly find out. It is more informative
to watch
the principal variation change over time. As it goes deeper the
sequence always
get better and better, and when you see that you know that increased thinking
will imrove the principla variation forever. This does not mean that always
plays the best move in the end, but it means that algorithm recursively
improves the analysis all over the tree and only a serious bug can stop that
from happen.
Note that MC-programs do not have high level concepts programmed into
them. Thus
you will alwyays be able to say "it played X because it does not understand Y"
whenver it plays a bad move but this is wrong. Correct is: "it played X
because
it did not search deep enough".
-Magnus
Quoting terry mcintyre <[EMAIL PROTECTED]>:
From: Don Dailey <[EMAIL PROTECTED]>
On Mon, 2007-01-22 at 03:43 +0100, alain Baeckeroot wrote:
The few games i played against mogobot on 19x19 shows that it does not
"know" overconcentration. And i can safely bet that increasing
thinking time will not solve this,
By definition, a scalable program can solve all problems so your
statement is incorrect.
Is Computer Go infinitely scalable? Is Mogobot, in particular, scalable?
Most programs reach scalability limits at some point. If the concept
of "overconcentrated shape" is not inherent in a given program, it
might take a vast number of simulations to compensate for that lack.
I suspect that monte carlo will need to be merged with some method of
learning patterns and concepts to play higher-level Go in human-scale
timeframes.
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
