Learning possibilistic graphical models from data

Résumé

This work fits within the framework of learning possibilistic networks, the possibilistic counterpart of Bayesian networks, which represent an interesting combination between possibility theory and graphical models. This thesis presents two major contributions. The first one consists on proposing a validation strategy for possibilistic networks learning algorithms. This strategy proposes a sampling process to generate imprecise datasets from theses models and two new evaluation measures. Our second contribution consists on proposing a global approach to learn the structure and the parameters of possibilistic networks. We propose a possibilistic likelihood function to learn possibilistic networks parameters and to define a new score function used to learn the structure of these models. A detailed experimental study showing the feasibility and the efficiency of the proposed methods has been also proposed.

Citer ce document

Accès au document

Ce document est hébergé sur son dépôt institutionnel d'origine.

Auteur(s)

Statistiques

Consultations : 2

Téléchargements : 0