Re: [computer-go] inside and outside influence

[tesuji@xxxxxxxxx]

Yes Bruno, I know your article. However, I wasn't aware it dealt with the
conceptual difference between 'inside' and 'outside' influence. I may be
mssing something in your paper. My definition of territory and eye-space
is different than your erosion function, but the end-result is similar. Or
maybe mine could be seen as a different kind of erosion function of only
two steps. My influence function is very similar to the Zobrist algorithm.
It does 8 iterations (called dilation in your paper, I looked it up to
refresh my memory) and includes the concept of 'dame-points' which always
get a zero value.

Over time I've come to recognise the importance of making an extra
classification of territory which I call 'inside' and 'outside'. Where
'inside' territory is much more valuable of course than outside influence.
A small number of points enclosed in a corner is often worth more than a
large area influence by the wall of the opposing color around the corner.
Until the outside influence threatens to become real territory of course,
which is exactly what this method attempts to detect. This happens when
points end up being 'inside' but not as 'territory' in the original
computation. This is the essence of a 'moyo'.

I don't have the original post at my disposal at the moment (although I
may still have it in my mailbox at home somewhere.) But what the idea came
down to is to treat the influence as mountainous area. To determine
whether a point belongs to the 'inside', a ball starts at that point and
rolls down-hill. When the ball comes to a rest on a 'neutral' point
(influence zero or opposite color) it was 'outside'. When the ball stops
at a local minimum (an elevation) the original point was 'inside'.

The actual algorithm needed a bit of tweaking to make it actually produce
useful results. Mathematically this makes the idea maybe less interesting,
but it produced the desired results in practice. The advantage was that I
managed to program it by a doing a very simple calculation, once for each
empty point on the board. This was based by work originally started by
Charlie Carroll, who worked for me at the time.

When I have a little more time I might make some diagrams and post them
with the algorithm, it will illustrate the idea much better than plain
words. The power of the algorithm is that it recognises territorial
frameworks, encouraging to make extensions from a wall and recognising how
to build a moyo. And it does this in a way that appeals (visually) to me
as a (6-dan) player, it almost never seems to come up with 'strange'
results that only a computer can 'see'.

And I think that's just the start. In a similar way there seems to be a
possibility to help determine the level of being 'enclosed' of a group by
the opposing color. And possibly detect escape-routes.

    Mark Boon

P.S. In another post I someone stated that in computer Go 95% of the time
is spent in search. Without any qualification this seems a strange
statement, one could probably say of the same program it spends 95% in
evaluation. Just depends on your point of view.


> Hi Mark,
>
> I also think that, unfortunately, there is a few interest on board
> static analysis on this list.
>
> I have been interested in the difference between 'inside' influence
> (territory) and 'outside' influence several years ago. My proposal was
> based on mathematical morphology explained in:
> http://www.math-info.univ-paris5.fr/~bouzy/publications/Bouzy-IJPRAI.pdf
> You surely know it. However, I post it to the list to avoid the "zero"
> interest on this crucial topic...
>
> Bruno
>
>
>
> tesuji@xxxxxxxxxxxxxxxxx a icrit:
>> There's also still a great deal to be gained in board-analysis. I
>> haven't
>> been following this mailing-list until recently again, but it always
>> strikes me that not much attention is given to new ways of gathering
>> information about the position in a static way. A few years ago I posted
>> an idea that I worked out about the concept of 'inside influence' and
>> 'outside influence' (also called 'territory' opposed to 'influence'). I
>> found a very elegant algorithm to support the idea too, but there seemed
>> to be zero interest for it. Where I think it was one of the most
>> exciting
>> things because it suddenly makes visible the difference between real
>> territory and a 'moyo', or potential territory.
>
>
> --
> ---------------------------------------------------------------------
> 	Bruno Bouzy
> 	C.R.I.P.5 - UFR de mathematiques et d'informatique
> 	Universite Rene Descartes (Paris V)
> 	45, rue des Saints-Peres 75270 Paris Cedex 06 FRANCE
> 	tel: (33) (0)1 44 55 35 58 fax: (33) (0)1 44 55 35 35
> 	e-mail: bouzy@xxxxxxxxxxxxxxxxx
> 	http://www.math-info.univ-paris5.fr/~bouzy/
> ---------------------------------------------------------------------
>
>
> _______________________________________________
> computer-go mailing list
> computer-go@xxxxxxxxxxxxxxxxx
> http://computer-go.org/mailman/listinfo/computer-go
>

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://computer-go.org/mailman/listinfo/computer-go