Two-year Post-doctoral position in statistics and decision theory Title: Model selection for control of partially observed spatial processes
Host team: The Biometry and Artificial Intelligence (BIA) laboratory in Toulouse is part of the French National Institute for Agricultural Research (INRA). The post-doc position is opened in the team Modelling of Agro-ecosystems and Decision (MAD) of the BIA lab. This team is composed of researchers in Artificial Intelligence and in Statistics. Its goal is to develop and apply mathematic and computer science methodologies contributing to the analysis of agricultural, forest and ecological systems and their management. (more at http://mia.toulouse.inra.fr/ →Equipes →MAD). Advisors: Dr. N. Peyrard (peyrard-AT-toulouse.inra.fr) and Dr. I. Chadès (chades-AT-toulouse.inra.fr) (more information at http://mia.toulouse.inra.fr/index.php?id=111). Keywords: Spatial statistics, partially observable Markov decision process, modelling, management of ecosystems Description: The MAD team aims at developing original decision-theoretic tools to solve complex optimization problems encountered in the management of agricultural and ecological systems. Lately, MAD has developed strong competences on Markov Decision Processes (MDP) methods. MDP are discrete-time stochastic models of controlled temporal processes, well adapted to the modelling of sequential decision problems under uncertainty. The work proposed is at the interface between Statistics and Decision Theory. The subject concerns decision problems that involve partially observable states structured by spatial constraints. Within this framework, important but yet unexplored statistical issues are model estimation and selection. Indeed, classical statistical methodologies do not take into account the controlled aspect of the model. For example, a good model regarding data fitting can perform poorly regarding the optimization problem. The objective of the project is to analyse the suitability of model selection methods and to develop an approach dedicated to partially observed and spatial MDP optimisation. This methodological part will be applied for validation to one of the applications in ecosystem management studied within the MAD team. Qualification and competences expected: Candidates should have a Ph.D. (or expect to receive one by February 2008) in the field of Statistics or Decision Theory. They should have a strong background in statistical processes (hidden Markov chain/ hidden Markov random fields, model selection), as well as competences in stochastic optimization techniques and programming. Ability to write scientific papers in English is required. Contract description: The open position is a 2-year full time position, starting in March 2008 (with some flexibility). Salary will be around 2100 Euros/month. -- Dr Iadine Chadès PhD in Artificial Intelligence Research Scientist INRA [EMAIL PROTECTED] http://mia.toulouse.inra.fr/index.php?id=57 Tel: +33(0)5 61 28 50 64 Fax: +33(0)5 61 28 53 35 Unité de Biométrie et Intelligence Artificielle, BP 27 31326 CASTANET-TOLOSAN cedex, FRANCE http://mia.toulouse.inra.fr/ _______________________________________________ uai mailing list uai@ENGR.ORST.EDU https://secure.engr.oregonstate.edu/mailman/listinfo/uai
