Graphical representations. Independence
Graphical representations of uncertain knowledge has been developed with Bayesian nets which exhibit conditional independence relations.
Possibilistic counterparts of Bayesian nets have been defined, using two different types of conditioning (see Possibility theory and related representations of uncertainty).
Anytime propagation algorithms have been proposed for possibilistic nets.
Various notions of independence between variables which can be defined in a qualitative uncertainty setting have been studied. Independence may refer to the idea of decomposing a relation without losing information, or to the fact that learning that a variable becomes true does not modify our belief regarding another variable.
contacts: S. Benferhat, J. Lang, J. Mengin.
some recent publications
- N. Ben Amor, S. Benferhat, D. Dubois, H. Geffner, H. Prade. Independence in qualitative uncertainty frameworks. Proc.7th Int.Conf.on Principles of Knowledge Representation and Reasoning (KR 2000), Breckenridge, Morgan Kaufmann, 235-246, 2000.
- S. Benferhat, D. Dubois, S. Kaci, H. Prade. Graphical readings of possibilisitc logic bases. Dans: Proc.of the 17th Conference Uncertainty in Artificial Intelligence (UAI01) , Seattle, Aug. 2 - 5, Morgan Kaufmann Publishers, 24-31, 2001 .
- J. Lang et P. Marquis. Complexity results for independence and definability in propositional logic. Proceeedings of KR'98, Morgan Kaufmann, 356-367, 1998.
- N. Wilson et J. Mengin. Embedding logics in the local computation framework, Journal of Applied Non-Classical Logics, Vol. 11, n° 3-4, p. 239-267, 2001.
