Output details
11 - Computer Science and Informatics
Teesside University
Improved use of partial policies for identifying behavioral equivalence
<22> This paper presents the most efficient and effective algorithm for solving interactive dynamic influence diagram (I-DID) by identifying behavioural equivalence in an incremental way. Different from the previous set of algorithms (Zeng, JAIR 2012), the technique allows the comparison of agents’ behaviour in an asymmetric way. This is important since it not only saves a large amount of computational memory in the comparison process, but also helps further compress the model space of I-DID thereby providing scalable solutions. This line of research offers a solid foundation for developing more scalable solutions to I-DID and enables practical applications of I-DID.