Re: some ideas

[Mika Kojo <mkojo@xxxxxx>]

Antti Huima writes:
> On Mon, 3 May 1999, Heikki Levanto wrote:
> 
> > Henrik Rydberg (rydberg@xxxxxxxxxxxxxxxxx) wrote in lsd.compgo:
> > 
> > : As opposed to for instance Backgammon [1], Go is strictly
> > : deterministic, leading to some problems when applying
> > : algorithms such as temporal difference learning (TD).
> > 
> > I think the problem with go is not so much the determinism, but the
> > difficulty in evaluating positions. 
> 
> I could second Rydberg's opinion in that the reason why TD succeeds for
> Backgammon is the stochastic nature of the game. The probabilistic
> component smoothes the state space so that indeed there are very `similar'
> positions and it is possible to generalize game experience. Contrary to
> that, go is very chaotic in the sense that minor changes in a board
> position can totally change its evaluation. I think this is the main
> reason why neural-net based approaches have failed. Neural nets are good
> for learning continuous functions but not chaotic ones.

Perhaps one can find a representation of a position that makes the
situation look less chaotic? Certainly Go is not entirely random game,
and there must be a way to look at the game that gives some sense of a
"good move" (I guess this has been proven empirically :). Just giving
enough hints for the network, in form of expert knowledge, might be
enough to remove most of the chaos.

It might be, that the obstacle in neural computing as applied to Go
is finding such good representations.



Mika Kojo
SSH Communications Security, Ltd.