Re: computer-go: Quiescence (was Re: Temporal Difference Learning)

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

At 02:45 PM 7/16/2003 -0700, you wrote:


> This raises another question:  is there a good cheap criterion for 
> quiescence in Go?

I don't know of any.  Many Faces treats a position as quiet if all groups 
are stable, or
if adding a move to the unstable groups leaves them unstable (indicating an 
error in
the evaluation function).  In practice this means that during quiescence 
search Many Faces tries moves to
attack or defend unstable groups and rejects moves that change the 
evaluation by less than the
value of a typical big move.

David Fotland


> > I don't think that it is a bad idea to use an evaluation
> > function that was trained with TD and 1-ply look-ahead
> > later in deeper searches. I am doing this with NeuroGo on
> > the 9x9 board and doing a 3-ply full-board search makes it
> > significantly stronger.
> Presumably the evaluation function has learned some intuitions about life, 
> influence, and territory.
>
>
> Peter Drake
> Assistant Professor of Computer Science
> Lewis & Clark College
> http://www.lclark.edu/~drake/