Artificial Neural Network Architecture for Solving the Double Dummy Bridge Problem in Contract Bridge

Abstract: Card games are interesting for many reasons besides their connection with gambling. Bridge is being a game of imperfect information, it is a well defined, decision making game. The estimation of the number of tricks to be taken by one pair of bridge players is called Double Dummy Bridge Problem (DDBP). Artificial Neural Networks are Non – Linear mapping structures based on the function of the human brain. Feed Forward Neural Network is used to solve the DDBP in contract bridge. The learning methodology, supervised learning was used in Back – Propagation Network (BPN) for training and testing the bridge sample deal. In our study we compared back – Propagation algorithm and obtained that Resilient Back – Propagation algorithms by using Hyperbolic Tangent function and Resilient Back – Propagation algorithm produced better result than the other. Among various neural network architectures, in this study we used four network architectures viz., 26x4, 52, 104 and 52x4 for solving DDBP in contract bridge.

Keywords: BPN, Contract Bridge, Back – Propagation Algorithm, Resilient Back – Propagation Algorithm, Hyperbolic Tangent function.