Actualités
  Période
semaine
ou mois
aucune période


  Thèmes:
Thèses
Colloques
Autres
Groupes
Tous les thèmes

   
  Liens:
Voir les thèses

Voir les colloques

Voir les autres événements

Voir la page des groupes

Accéder à l'intranet

Intitulé:   
Reachability in MDPs: Refining Convergence of Value Iteration
du groupe Modélisation et Verification

Date 2015-03-05  11:00-12:00
TitreReachability in MDPs: Refining Convergence of Value Iteration 
RésuméMarkov Decision Processes (MDP) are a widely used model including both non-deterministic and probabilistic choices. Minimal and maximal probabilities to reach a target set of states, with respect to a policy resolving non-determinism, may be computed by several methods including value iteration. This algorithm, easy to implement and efficient in terms of space complexity, consists in iteratively finding the probabilities of paths of increasing length. However, it raises three issues: (1) defining a stopping criterion ensuring a bound on the approximation, (2) analyzing the rate of convergence, and (3) specifying an additional procedure to obtain the exact values once a sufficient number of iterations has been performed. The first two issues are still open and for the third one a « crude » upper bound on the number of iterations has been proposed. Based on a graph analysis and transformation of MDPs, we address these problems. First we introduce an interval iteration algorithm, for which the stopping criterion is straightforward. Then we exhibit convergence rate. Finally we significantly improve the bound on the number of iterations required to get the exact values. 
Lieu075 
OrateurBenjamin Monmege 
Emailbmonmege@ulb.ac.be 
UrlULB 



Aucun document lié à cet événement.

Retour
Retour à l'index