RE: computer-go: Fast scoring program?

["Mark Boon" <tesuji@xxxxxxxxx>]

Yes, I was actually also going to suggest trying something more simple. I
just don't believe a 'simple' neural net will be able to learn to play Go
any more than randomly mixing some chemicals and wait for life to
spontaneously come to be.

Even atari-Go I have doubts whether it can be learned by a simple NN because
the process of deciding good or bad rely on very exact information, whereas
I've always had the impression NNs are more suitable for fuzzy problems.

I was wondering if an influence algorithm could be done using a NN. Ideally,
an influence algorithm calculates the probability of a certain point that it
becomes territory for a certain color. As feedback one can use the end of
the game, when all points have been decided. Would this be a feasible idea?

    Mark

> -----Original Message-----
> From: owner-computer-go@xxxxxxxxxxxxxxxxx
> [mailto:owner-computer-go@xxxxxxxxxxxxxxxxx]On Behalf Of Matthew Corey
> Brown
> Sent: Friday, August 17, 2001 2:21 AM
> To: computer-go@xxxxxxxxxxxxxxxxx
> Subject: Re: computer-go: Fast scoring program?
>
>
> On Thu, 16 Aug 2001, John Tromp wrote:
>
> > why not try set the nets to an easier task, where the first to
> > capture opponent's stones wins the game? that avoids both passing
> > and scoring problems. this is a popular way to introduce go to
> beginners,
> > so it's only natural to use it for nets which are the supreme
> beginners...
>
> Actually thats a really good simple fitness function. Thanx for the idea..
>
> There is beauty in simple. It predijuices the nets less.
>
> >
> > also try limiting yourself to 5x5 or 7x7 boards to accelerate the
> > learning process.
> >
> >
>
> Ahh but if it gets good i want to put it on the ladder so 9x9 is size i
> need
>
>
> Matthew Corey Brown                             bromoc@xxxxxxxxxxxxxxxxx
>                   Happiness is a dry place to live.
>
>