Anytime propagation algorithm for min-based possibilistic graphs

This paper proposes a new anytime possibilistic inference algorithm for min-based directed networks. Our algorithm departs from a direct adaptation of probabilistic propagation algorithms since it avoids the transformation of the initial network into a junction tree which is known to be a hard problem. The proposed algorithm is composed of several, local stabilization, procedures. Stabilization procedures aim to guarantee that local distributions defined on each node are coherent with respect to those of its parents. We provide experimental results which, for instance, compare our algorithm with the ones based on a direct adaptation of probabilistic propagation algorithms.