Re: Fuseki and Joseki Database with Neural Network

["Magnus Persson" <magnus.persson@xxxxxxxxxx>]

>  I'm not an expert on ANN. But I'm not sure joseki is best solved with ANN.
> Because my experience with ANN is that it is not exact. But ANN can "guess"
> where it is not sure. So maybe ANN is good for choosing fuseki and maybe
> choose joseki depending on the fuseki. But to store a joseki sequence I
> think normal datamodels is better than ANN.

To explain why datamodels is better I think the folowing is a good 
why of thinking.

At each step in a sequence of forced moves in a joseki, there are 
many alternative moves  that look good as local tactical shapes, 
but nethertheless the joseki only allows for one correct move - all 
the other alternatives are mistakes.

I think that ANN has the potential to discriminate those moves that 
seems to be alternatives to the human eye from those that are never 
considered at all. But an ordinary ANN however properly trained, will 
not be able to make the correct choice between these alternatives. To 
do this one has feed the ANN with information about higher concepts 
of go theory, which in itself is what we wanted the ANN to learn 
without our help...

Playing correct joseki and choosing the right variations is very 
difficult. If ANN's could do that well then humans would do it even 
more easily.

My approach to computer go is to avoid fuseki. One has accept the
fact that the program do not understand enough to play well in the
fuseki. I think one should concentrate on making programs that are
strong at the endgame and middle game fighting. If one is
successfull there, the program will play a good fuseki as a
consequence of the true playing strength of the program. 

Trying to force a program to play "correct" joseki is often just a
way of hiding the weaknesses of the program - and might in the long
run be counterproductive to the development of programs that can
play well to the end of the game and eventually win. 

Magnus Persson, Swedish 2 dan
--
Magnus Persson
magnus.persson@xxxxxxxxxxxxxxxxx
http://www.docs.uu.se/~magnuspe
Department of psychology, Uppsala University