RE: [computer-go] Pattern Matcher

["Mark Boon" <tesuji@xxxxxxxxx>]

OK, maybe I was not clear enough, so I'll try to answer your questions that
I thought were obvious.

> -----Original Message-----
> From: computer-go-bounces@xxxxxxxxxxxxxxxxx
> [mailto:computer-go-bounces@xxxxxxxxxxxxxxxxx]On Behalf Of Vincent
> Diepeveen
> Sent: Monday, November 08, 2004 16:47
> To: computer-go; computer-go
> Subject: RE: [computer-go] Pattern Matcher
>
>
> You wonder how big the speedup will be.
>
> My first question is: "how many patterns, 8 million auto
> generated patterns?"

Several hundreds, possible extended to thousands or ten-thousands when they
get generated instead of hand-coded. I expect many different 'sets' of
patterns though. Right now I have a test-sample of 356 patterns that I can
match half a million times per second to a single location. I expect on
average to have to match a few dozen locations per move, so say 100 uSec.
per move on a 900Mhz Athlon.

(On a completely different note, I was even philosophising about
meta-patterns. Patterns that tell me where to look for patterns, so I reduce
the number of locations where I need to look for them. Maybe these
meta-patterns can be automatically generated for a set of patterns. Or is
this too much off the wall?)

> My second question is: "to what do you want to compare it too,
> hand written
> C code?"

I want to compare it to (hand-written) Java code (or C-code if you like)
that takes a board-position and the datastructure that holds the
decision-tree, and recursively matches the decision-tree to the board.

> The third question is: "what does a pattern actually do?"
> Like does every pattern that occurs in a given board position X give a
> 'bonus' for a move 'x' to be tried first and the move x with the highes
> score is the move we go try first?
>

This is irrelevant. I just need to know which pattern can be found where.

> Or does each pattern have its own values?
>
> Of course it has been proven already that when pattern amount keeps
> growing, that auto tuners do not know how to assign score to patterns when
> compared to the superiority of certain programmers there (not necessarily
> good players)

This is an altogether different issue. Right now I'm just interested in
locating patterns.



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