In article <komodoE9p6Dx.3Lx@netcom.com>, komodo@netcom.com (Tom Johnson) writes:
|> As to the ability to learn, the top backgammon playing programs were all |> trained on some form of neural network and 'learned' by making error |> corrections in response to winning and losing moves.This method does not work |> well in chess for many reasons; the main one |> being that chess is not a deterministic game, ie, doesn't have to have an |> ending.
I was with you until this point. Chess is deterministic, in the generally accepted meaning of the term. Backgammon is not, because it has a chance component. Whether or not the game must terminate is not really relevant to this point, but in fact the rules of chess do force it to terminate, while AFAIK a backgammon game could go on forever (although the probability of this happening would be 0).
Robert Israel israel@math.ubc.ca Department of Mathematics (604) 822-3629 University of British Columbia fax 822-6074 Vancouver, BC, Canada V6T 1Y4
