Improving MACS thanks to a comparison with 2TBNs

Gourdin, Thierry

Sigaud, Olivier
LIP6 - Pôle IA
Université Paris 6
8, rue du capitaine Scott F-75015 Paris, France
email: olivier.sigaud@lip6.fr   

Wuillemin, Pierre-Henri
LIP6 - Pôle IA
Université Paris 6
8, rue du capitaine Scott F-75015 Paris, France
email: pierre-henri.wuillemin@lip6.fr   home: www-desir.lip6.fr/~phw

Gourdin, Thierry and Sigaud, Olivier and Wuillemin, Pierre-Henri (2004) "Improving MACS thanks to a comparison with 2TBNs". In Proceedings of Genetic and Evolutionary Computation Conference, pp. 810--823, Springer-Verlag.

Factored Markov Decision Processes is the theoretical framwork underlying multi-step Learning Classifier Systems research. This framework is mostly used in the context of Two-stage dynamic Bayesian Networks, a subset of Bayesian Networks. In this papper, we compare the Learning Classifier Systems approach and the Bayesaisn Networks approach to factored Markov Decision Problems. More specifically, we focus on a comparison between MACS, an Anticipatory Learning Classifier System, and Structured Policy Iteration, a general planning algorithm in the context of Two-stage Bayesian Network. From that comparison, we define a new algorithm resulting from the adaptation of Structured Policy Iteration to the context of MACS. We conclude by calling for a closer communication between both research communities.

@InProceedings{,
  author = {Gourdin, Thierry and Sigaud, Olivier and Wuillemin, Pierre-Henri},
  title = {Improving MACS thanks to a comparison with 2TBNs},
  booktitle = {Proceedings of Genetic and Evolutionary Computation Conference},
  year = {2004},
  pages = {810--823},
  publisher = {Springer-Verlag}
}