Probabilistic neural network playing a simple game

0411134 - UTIA-B 20030121 RIV IT eng C - Konferenční příspěvek (zahraniční konf.)
Grim, Jiří - Somol, Petr - Pudil, Pavel - Just, P.
Probabilistic neural network playing a simple game.
Florence: University of Florence, 2003. In: Artificial Neural Networks in Pattern Recognition. Proceedings. - (Marinai, S.; Gori, M.), s. 132-138
[IAPR TC3 Workshop 2003 /1./. Florence (IT), 12.09.2003-13.09.2003]
Grant CEP: GA ČR GA402/01/0981; GA ČR GA402/03/1310; GA AV ČR KSK1019101
Výzkumný záměr: CEZ:AV0Z1075907
Klíčová slova: probabilistic neural networks * finite mixtures * EM algorithm
Kód oboru RIV: BB - Aplikovaná statistika, operační výzkum

The goal of the paper is to design a probabilistic neural network playing a simple two-player game "Tic-Tac-Toe". The game is considered as a problem of a repeating evaluation of preferences of possible moves. Assuming the probabilistic neural network in the role of the evaluation function we can solve the problem by estimating the probability distribution of advantageous moves. The unknown distribution is estimated in the form of a finite discrete mixture of product components by means of EM algorithm.
Trvalý link: http://hdl.handle.net/11104/0131221