Re: [computer-go] Bouzy/Zobrist algorithms

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

Bruno Bouzy wrote:
> Erik van der Werf a écrit :
>
> > Peter Drake wrote:
> >
> > > Since the erosion happens after the dilation, does that mean that the 
> > > values for the occupied points are never altered by either operation?
> >
> >
> > Interesting question. I think that, at least for the (5,21) version, 
> > some extremely weak stones may be removed by the erosions. (The values 
> > for occupied points can of course change because otherwise there would 
> > be no point in using large values.) I'm not sure if this was Bouzy's 
> > intention, but it might actually help to improve performance. Maybe 
> > Bruno can shed some light on this?
> >
> > Erik
> > _______________________________________________
> > computer-go mailing list
> > computer-go@xxxxxxxxxxxxxxxxx
> > http://www.computer-go.org/mailman/listinfo/computer-go/
>
> You go further than me. 
I'm sorry, but I don't think so.


> The "occupied points" (points occupied by a 
> stone) were not the aim of the algorithm. The initial value (say -+64) 
> is supposed to be high enough to keep these points with their initial 
> sign. 
It is not high enough for d=5,e=21. With 21 erosions you can maximally 
subtract 84. If the dilations had insufficient effect this means that 
some stones may be gone after the erosions. (At least if I understand 
your description correctly. Or did you use -+85 for the d=5, e=21 version?)

Erik


PS Just try something like:

. . .
. O #
. # .
in the center of a sufficiently large board

maybe even this one will work:
. . . . #
. . O . .
# . . . .
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/