Selecting Learning Algorithms for the Design of Virtual Players in Natural Resources Management Platforms
Guillaume M Uller, Jaime S. S Ichman, Escola Polit · 2009
This paper focuses on the problem of designing Virtual Players for a particular type of serious game, in the domain of Natural Resources Management. Indeed, in some cases, it is impossible to play a particular game because there is not enough participants. Designing and implementing Virtual Players, that would stand in for some of the lacking human players, would allow to circumvent this limitation. In order to get realistic substitute players, it has been proposed to design the Virtual Players by learning actual human behaviours based on traces of games in which they are involved. However, there exists plenty of learning algorithms, with very different requirements and properties and no a priori best algorithm. This paper addresses more particularly the problem of selecting learning algorithms for the design of Virtual Players in Natural Resources Management “ABPS” platforms. Its main contributions are: a list of criteria for practically comparing the learning algorithms and the selection of some algorithms to be used in the design of Virtual Players for Natural Resources Management “ABPS” platforms.