Re: [computer-go] Pattern matching - rectification & update

["Frank de Groot" <frank@xxxxxxxxxxxxx>]

From: "Mark Boon" <tesuji@xxxxxxxxxxxxxxxxx>
Subject: RE: [computer-go] Pattern matching - rectification & update



> Do you mean inefficient in algorithm, or inefficient in implementation?
> Do you mean tactical in terms of life-and-death or tactical like in
ladders,
> geta, etc...?


I mean that I do not understand why they strive for a guaranteed way to
solve a problem.
Isn't it better to be a trillion trillion times faster and be only 90% sure
you solved the problem?

ButI am a boneheaded person.
I never understood the Theory of Special Relativity and nobody was able to
explain it to me (just like with all those fancy search techniques).

When I finally read Einsteins own book about it, it did not give me
understanding, on the contrary, I said: "So this whole theory is nonsense,
as Einstein confuses the Doppler effect of light with the actual slowing
down of time".

:)

> I agree, millions of nodes focus on quantity instead of quality. And I
> consider Go to bemore of a qualitative game.


Oh. I have no idea. It looks rather boring to me, compared to Go programming
:)


> I think we came a long way of making a life-and-death module that does
what
> you're looking for. It was based on whole-board evaluation combined with
> proof-number search. Simple problems would be solved in hundreds of nodes,
> complex ones in thousands. To 'solve' carpenter-square took maybe
> ten-thousand, but that even stumps your average 6d amateur.

Well it is interesting to work on such a thing.

> The search-algorithm is simple, the evaluation is not. But I agree it's
> doable.

I don't understand the theory behind the search algorithms (except
Alpha-Beta with killers)
I want to make something much simpler that does exactly what a human does.

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