RE: computer-go: perfect play : a new fractal algorithm

["Cuijpers, P.J.L." <P.J.L.Cuijpers@xxxxxx>]

Indeed I remember I have seen a similar technique before in image-processing
techniques. The movement of a biological cell was studied using so called
feature-fraction techniques (if I remember correctly). This is similar, but
more general, than the proposed fractal-algorithm method. Nevertheless we
remain with the problem of defining the right features.

> -----Original Message-----
> From: Vlad Dumitrescu [mailto:vladdu@xxxxxxxxxxxxxxxxx]
> Sent: vrijdag 29 september 2000 10:23
> To: computer-go@xxxxxxxxxxxxxxxxx
> Subject: Re: computer-go: perfect play : a new fractal algorithm
> 
> 
> my first impression: interesting approach!
> 
> the big problem seems to be how to reduce the board size 
> without losing
> importnt information. If a reduced group becomes dead 
> (because there isn't
> enough space to make it alive), then the reduction fails...
> 
> one way to handle this might be using a method presented here 
> for som months
> ago - I'm not sure I remember correctly -- building a 
> "feature tree" where a
> string of stones or liberties becomes just a node... if the 
> reduced board
> position has the same feature tree, then it might be close 
> enough to the
> original one.
> 
> Why do you say the branching factor is reduced from 361 to 4? it is so
> directly at the 19x19 level, but you have to search the 
> reduced boards too,
> thus there are more moves to look into before being able to 
> choose one of
> the 4...
> 
> new approaches are always welcome! :-)
> /Vlad
>