simonof@aplcen.apl.jhu.edu (Simonoff Robert 301 540 1864) (10/25/90)
I am interested in hearing from anyone who has experience with training neural networks on playing games. I am specifically interested in automatic learning of game playing strategy - not simply just giving the network a series of moves to memorize. I have a theory in this field that I would like to bounce off interested people and see if it is original. Thanks. Bob Simonoff -- *********************************************************** Bob Simonoff simonof@aplcen Johns Hopkins University
rob@modus.sublink.ORG (Roberto O. Buratti) (10/30/90)
In article <1990Oct24.213603.23176@aplcen.apl.jhu.edu>, simonof@aplcen.apl.jhu.edu (Simonoff Robert 301 540 1864) writes: > ...I am > specifically interested in automatic learning of game > playing strategy - not simply just giving the network > a series of moves to memorize. I have a theory in this field > that I would like to bounce off interested people and see if it is > original. I am not experienced with NN, but I too have developed an (interesting?) idea in this field. Suppose you have a 2-player perfect-info game such as chess, ttt, etc... You can then use a net as move-generator / evaluation-function in a canonical alpha-beta search. I would like to discuss this idea with experts and read some paper, but at the moment I have no bibliografy to send to you. Sorry if I can't help more than this... Please forward to me any interesting paper. Rob -- _ _ __ Via Aleardo Aleardi, 12 - 20154 Milano (Italy) | | | _ _| (__ PHONE : +39 2 3315328 FAX: +39 2 3315778 | | |(_)(_||_|___) Srl E-MAIL: rob@modus.sublink.ORG ______________________________ Software & Services for Advertising & Marketing