Re: computer-go: Neural Nets: suggesting and evaluating

[Erik van der Werf <E.vanderWerf@xxxxxxxxxxxxx>]

Peter Drake wrote:

> > > 2) A lot of time is wasted recomputing the same information over and 
> > > over again.
> >
> > If it's really the same information why can't you just store it in a 
> > cache?
>
> Well, it's not EXACTLY the same information, but I imagine the hidden 
> units for a network looking at a window around d3 are computing a lot 
> of the same information as those in a network looking at a window 
> around d4. 
Yes, that is a problem.

Just a thought: Maybe you can start training on a per point basis (which 
is easier). Then when the local predictions are good enough you could 
combine each local predictor in a full board predictor network which can 
then be pruned (and possibly retrained) untill enough redundant hidden 
nodes are gone.

Best,
Erik