[computer-go] when to stop searching

[Don Dailey]

When I speak of true selectively I actually mean programs 3 or 4 by
your definition, it's just a matter of semantics.   I mean a program
that doesn't consider every move and once it's made the decision not
to consider a move will never revisit or modify that decision.

Although a "true selective" program is self-limiting, you can build
a program of any arbitrary strength provided you know what you are
doing.   It's a perfectly valid approach.

- Don



On Sun, 2007-03-11 at 22:59 -0700, David Fotland wrote:
> I would add 4.  The program tries to identify good moves, and only tries 
> moves that it thinks might be good.  If it is goal-directed, the good moves
> are good for a reason, and if the reason is not satisfied, they are
> discarded.
> 
> This is the way Many Faces works.  It's very similar to 3, but it's a
> different 
> way of thinking about the problem (adding good moves rather than deleting
> bad ones).
> 
> You are correct that this approach is inadmissible, and is self limiting.
> 
> I like it because when I add new knowledge I know I'm making the program
> stronger.  
> I don't like tuning parameters or algorithms and then playing hundreds of
> games
> to get statistically significant data on the better value.  I'm not saying
> my approach
> better, just that I prefer it :)
> 
> David
> 
> > 
> >   3.  selective in the "true" sense.  Such a program tries to 
> >       identify bad moves and prune them from the tree, but they
> >       are pruned permanently.  NO matter how deep or long you 
> >       search they will never be considered.   
> > 
> > I think "true" selective programs, unless the pruning 
> > criteria is fully admissible,  is self limiting.  You can 
> > probably build a strong program but you will be bound 
> > strictly to the quality
> > of your selective algorithm.   Such an algorithm would play
> > imperfectly even on an infinte speed computer.
> > 
> > UCT is admissible - it will ALWAYS find a winning move if you 
> > are in a winning position.
> > 
> > - Don
> > 
> > 
> > _______________________________________________
> > computer-go mailing list
> > computer-go at computer-go.org 
> > http://www.computer-go.org/mailman/listinfo/computer-go/
> > 
> 
>