Supervised Learning in Imperfect Information Game

Abstract

Bridge is an international imperfect information game played with similar rules all over the world and it is played by millions of players. It is an intelligent game, which increases the creativity with multiple skills and knowledge of human mind. Because no player knows exactly what moves other players are capable of making. It is viewed as an imperfect information game. It is well defined in particular; the scoring system gives a way of assessing the strength of any resulting program. Artificial Neural Networks (ANN) is trained on sample deals without any idea of game and used to the estimate the number of tricks to be taken by one pair of bridge players in the so-called Double Dummy Bridge Problem (DDBP). Supervised learning was used in Back-propagation neural network to take best tricks in Double Dummy Bridge Problem. The target data was trained and tested using Log Sigmoidal transfer function and Hyperbolic Tangent Sigmoid functions, in the intermediate layer of the network. The results of both the functions seem to be more convincing regarding the results of the game. In summary, the study described in this paper provides a detailed comparison between Log Sigmoid transfer function and Hyperbolic Tangent Sigmoid functions which were used to train and test the data

Keywords: Double Dummy Bridge problem, BPN, Supervised learning, Log Sigmoid transfer function, Hyperbolic Tangent Sigmoid function.