Belief merging and logic programming with stable model semantics
Representation of knowledge is a central aspect in artificial intelligence. Decisions of an intelligent agent must rely on informations which represent the world as correctly as possible. This difficulty to have a correct representation of the world is particularly important when dealing with a chan...
Main Author: | |
---|---|
Other Authors: | , , , |
Format: | Doctoral or Postdoctoral Thesis |
Language: | French |
Published: |
HAL CCSD
2009
|
Subjects: | |
Online Access: | https://theses.hal.science/tel-00529852 https://theses.hal.science/tel-00529852/document https://theses.hal.science/tel-00529852/file/These_Julien_Hue.pdf |
Summary: | Representation of knowledge is a central aspect in artificial intelligence. Decisions of an intelligent agent must rely on informations which represent the world as correctly as possible. This difficulty to have a correct representation of the world is particularly important when dealing with a changing world of information from multiple sources. In this thesis, we propose a method to perform syntactic merging in the case where no explicit priorities are expressed. This method, called Removed Sets Fusion (or RSF), is based on the search of subsets of formulas to remove to restore consistency. We performed an implementation of this method based on the translation of the fusion problem into a logic program with stable model semantics. We propose two different implementations: an adaptation of the smodels algorithm and another implementation based on the additional statements provided by Lparse/Gringo. We tested this last implementation: thanks to randomly generated belief profile and, in a second time thanks to the data from archeological survey coming from the European project VENUS. We then extended the Removed Sets Fusion in two directions. We study merging in the case where priorities are expressed between agents or between agents and beliefs. We also study the case were beliefs are expressed in terms of logic programs. L'acquisition et la représentation des connaissances est un aspect central dans le domaine de l'Intelligence Artificielle car une machine intelligente doit avant tout s'appuyer sur des informations représentant le monde de façon suffisamment précise. Cette difficulté à disposer d'une représentation correcte du monde est particulièrement importante lorsque l'on a à faire à un monde changeant ou à des informations provenant de sources multiples. Nous proposons dans cette thèse une méthode de fusion syntaxique de croyances dans le cas où les croyances sont représentées dans le cas où il n'existe pas de priorités explicites ni entre les agents, ni entre les croyances exprimées par les agents. ... |
---|