Buy GNU books!
|[ < ]||[ > ]||[ << ]||[ Up ]||[ >> ]||[Top]||[Contents]||[Index]||[ ? ]|
In GNU Go 2.6 extensive use was made of an algorithm from Bruno Bouzy's dissertation, which is available at: ftp://www.joy.ne.jp/welcome/igs/Go/computer/bbthese.ps.Z This algorithm starts with the characteristic function of the live groups on the board and performs `n' operations called dilations, then `m' operations called erosions. If n=5 and m=21 this is called the 5/21 algorithm.
The Bouzy 5/21 algorithm is interesting in that it corresponds
reasonably well to the human concept of territory. This
algorithm is still used in GNU Go 3.2 in the function
estimate_score. Thus we associate the 5/21 algorithm
with the word territory. Similarly we use words
moyo and area in reference to the 5/10
and 4/0 algorithms, respectively.
The principle defect of the algorithm is that it is not tunable. The current method of estimating moyos and territory is in `influence.c' (see section 16. Influence Function). The territory, moyo and area concepts have been reimplemented using the influence code.
The Bouzy algorithm is briefly reimplemented in the file `scoring.c' and is used by GNU Go 3.2 in estimating the score.
Not all features of the old `moyo.c' from GNU Go 2.6 were reimplemented--particularly the deltas were not--but the reimplementation may be more readable.
|webmaster||delorie software privacy|
|Copyright © 2003 by The Free Software Foundation||Updated Jun 2003|