Re: computer-go: Who has (read) the manuscript of Sei & Kawashima (2000)?

[Erik van der Werf <E.vanderWerf@xxxxxxxxxxxxx>]

Dear Shinichi Sei,

Thank you for putting the translation on-line, it's very interesting.

I still have some questions:

- Can you say anything on time? How long did it take to solve 4x4?

- Can you make an estimate of the number of nodes and/or the amount of
time that would be required to solve 5x5 (and 6x6)?

- Did you ever compare the performance of your move ordering heuristics
to domain-independent move ordering heuristics, such as the combination
of first the transposition move, then 2 Killer moves, and ordering the
rest by the history heuristic?

- Did you use iterative deepening?

- Your evaluation value was one of three values( win/defeat/draw ), did
you try heuristic values to guide the search? Did it take more nodes on
4x4? I guess that on larger boards heuristics, such as the differences
in points, may actually reduce the number of nodes by improving the move
ordering.

Best regards,
Erik



Shinichi Sei wrote:
> 
> Dear Erik,
> 
> I am Shinichi Sei.
> I put the paper on my page.
> Please refer to http://homepage1.nifty.com/Ike/katsunari/index_e.html
> 
> Sincerely,
> 
> ---
> Shinichi Sei : ssei@xxxxxxxxxxxxxxxxx
> http://homepage1.nifty.com/Ike/katsunari/index.html   (Japanese)
> http://homepage1.nifty.com/Ike/katsunari/index_e.html (English)
> 
> >Hi all,
> >
> >I'm looking for the following publication:
> >
> >S. Sei and T. Kawashima. A solution of Go on 4ラ4 board by game tree
> >search program. Manuscript, Fujitsu Social Science Laboratory, 2000.
> >
> >
> >It's been referenced to in Papers by a number of people, but all I could
> >find from the original was the following abstract:
> >
> >----------------------------------------------------------------------
> >IPSJ SIGNotes Game Informatics Abstract No.004 - 011
> >
> >
> >A Solution of Go on 4x4 Board by Game Tree Search Program
> >
> >Sei Shinichi, Kawashima Toshiaki
> >
> >Fujitsu Social Science Laboratory Ltd.
> >
> >We had made a search program to find a solution of Go on mini-board (4x4
> >or smaller) based on the Japanese Go rule. This paper describes the
> >solutions and the method of our program. As the result of executions, we
> >found that the solution were draw on 4x4 board, victory of black on 3x3
> >board, and draw on 2x2 board.
> >----------------------------------------------------------------------
> >
> >I'm especially interested in the type of search they applied. Was it an
> >alpha-beta with enhancements, a kind of PN-search, or something
> >original. It would be nice to have the publication, but since I expect
> >it to be written in Japanese, I'd be very happy if somebody who actually
> >read the paper could fill me in on the details.
> >
> >Thanks in advance,
> >Erik
> >