[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: computer-go: Programs learning to play Go



Dave Dyer wrote:
> 
>  My observation on NN based learning is that the few cases where
> they have worked well involve situations where choosing one of a
> few alternatives based on a few inputs is the mode.  The prototype
> success story is Backgammon.

Don't forget that this success story was still a combination of neural
networks AND search. 

>  Go doesn't really fit this model.  Throwing 361 inputs and outputs
> at a naive network, and expecting it to learn anything, is like
> throwing a pail of sand into an oven and expecting silicon chips
> to emerge.  Not fucking likely.

Your intuition might be a bit off here. A friend of mine who used
exactly this approach in chess middle-game positions had the amazing
result of predicting the optimal move 33% of the time! This ofcourse is
not enough to build a strong player, but saying that such a network
cannot learn anything seems a few bridges too far. 

> #1: if you must try a pure learning technique, start with a reduced
> domain; 3x3 or 4x4 is a good place to start.  Achieve percect play
> on 4x4, then work your way up.

Don't worry about perfect play, learn to predict reasonable moves, and
try to use them in some kind of search. (you might for example try this
approach for move ordering)

Erik