Asymmetric Game Trees using Dynamic Bayesian Networks

Student thesis: Master thesis (including HD thesis)

  • Esben Skov Pedersen
2. term, Computer Science, Master (Master Programme)
This project we improve mini max search by prediciting which move the opponent is most likely to take. This prediction is performed by using a dynamic bayesian network. This network has nodes that represents a number of abstract features ex- tracted from the game. The network is trained using EM-learning based on data from a number of comptuter players playing against each other. These computer players are defined as the weighted sum of the features mentioned before. These weights are found using a genetic approach. The system is able to improve perfor- mance of the worst strategies with- out increasing the search space.
Publication date2008
Publishing institutionInstitut for datalogi
ID: 14634038