Re: computer-go: Alpha-Beta Search

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 09:52 AM 11/18/99 +0000, you wrote:
>Vincent Diepeveen wrote:
>
>> After proof number search, some started working at conspiracy number
>> search.
>
> I believe it is the other way round. proof number is a double AND/OR
>whilst conspiracy is a single AND/OR tree.
>
>> I'm not very impressed by either proof number search nor conspiracy number
>> search.
>> 
>> Sure the ideas are good, but a very simple form of searching already
>> kicks hell out of them.
>
> References ?

Advances in ICCA there was a big article on it, 
and if i checkout the journal i probably see a similar
article again.

Basically see how a conspiracy number program like Conners at 180
processors plays.

Observing it for 3 tournaments now, it is seeing some lines that are
not interesting very deeply, yet some obvious losses either at itselve
or at the opponent it doesn't see if the eval doesn't predict the
moves very well.

I can't remember any other selective program that has gotten so far 
as Conners though, so that the program plays that strong definitely 
is a big victory already for Ulf.

Ask Ulf Lorenz, university Paderborn
for some (online?) sources or a big reference list. 
I'm sure he has one.

A major problem in CNS/PNS is how to apply hashtables efficiently?
The speedup you get with it in GO is quite huge!

>[snip interesting stuff]
>  cheers Paul.

Greetings,
Vincent
Vincent Diepeveen
diep@xxxxxxxxxxxxxxxxx

---
... en verder ben ik van ben ik van mening dat Dap het gesticht in moet