JEANTET Gildas

Supervision : Patrice PERNY

Co-supervision : SPANJAARD Olivier

Algorithmes pour la décision séquentielle dans l'incertain : optimisation de l'utilité espérée dépendant du rang et du critère de Hurwicz

In this thesis, we are interested in sequential decision problems under uncertainty. These problems concern situations of uncertainty where the decision maker has to make several decisions spread over time (i.e., establish a strategy). This problem is much studied in artificial intelligence, known as planning under uncertainty, because of its several applications in many fields (medical diagnosis, artificial players, autopilot, inventory management, ...). The economist community has provided many decision criteria for reasoning under uncertainty in order to compare strategies. However, the difficulties associated with their implementation leads in practice to use criteria less efficients in sequential decision problems. The use of performant criteria is indeed hindered by the lack of efficient algorithms in the computer science literature. The purpose of this thesis is precisely to tackle these algorithmics locks by providing algorithms for optimizing these criteria in sequential decision problems.

Defence : 09/27/2010

Jury members :

PERNY Patrice - Professeur à l'Université Pierre et Marie Curie (UPMC)
SPANJAARD Olivier - Maître de Conférence à l'Université Pierre et Marie Curie (UPMC)
BOUYSSOU Denis - Directeur de recherche CNRS à l'Université Paris-Dauphine (Paris IX) [Rapporteur]
FARGIER Hélène - Directeur de recherche CNRS à l'Université Paul Sabatier (Toulouse III) [Rapporteur]
CHRETIENNE Philippe - Professeur à l'Université Pierre et Marie Curie (UPMC)
GRABISCH Michel - Professeur à l'Université Panthéon-Sorbonne (Paris I)
SABBADIN Régis - Chargé de recherche INRA (Toulouse)

Departure date : 10/01/2012

2008-2012 Publications