Jacques Droulez

26 avril 2013

Collège de France, CNRS, UMR 7152, LPPA, Paris

Decision making under uncertainty : a quasimetric approach

invité par Manuel Vidal


Abstract : Making decisions under uncertainty is a major issue in a wide range of domains, including economics, artificial intelligence, cognitive sciences and motor control among many others. Basically, an agent has to choose a single or series of actions from a set of options, without knowing for sure their consequences. Schematically two main approaches have been followed : either the agent learns an action policy through experience (model-free approach), or the agent has already some knowledge, expressed in probabilistic terms, about the possible consequences of his decisions (model-based approach). In the latter case, finding the optimal action policy is often an intractable problem. This has led scientists to develop approximations or problem reformulations. Here we propose a new general approach, taking advantage of the intrinsic geometry of the state space. This method exhibits some interesting properties and may account for risky behavior frequently observed in humans. It also allows a substantial computational gain that could inspire new ways of solving control and decision problems.

CNRS logo université Aix Marseille logo | plan du site | mentions légales | contact | admin | intranet | intcloud |