Skip to Main content Skip to Navigation
Directions of work or proceedings


Souhila Kaci 1 Robert Mercer Matthias Thimm 2
1 SMILE - Système Multi-agent, Interaction, Langage, Evolution
LIRMM - Laboratoire d'Informatique de Robotique et de Microélectronique de Montpellier
Abstract : Many problems in AI (in reasoning, planning, learning, perception and robotics) require the agent to operate with incomplete or uncertain information. Uncertain Reasoning (UR) addresses these challenges by developing models building on both qualitative and quantitative uncertainty models for knowledge representation and applying them in diverse fields. This special issue collects the best papers from the special tracks on uncertain reasoning at the International Florida Artificial Intelligence Research Society Conference (FLAIRS) from 2014 and 2015 and selected papers from the general community. All submissions went through an additional reviewing phase to ensure a high quality for this issue. The special track on uncertain reasoning is the oldest track in FLAIRS conferences, running annually since 1996. The UR'2014 track at FLAIRS-27 was the 19th in the series while the UR'2015 track FLAIRS-28 was the 20th in the series. In “Hierarchical beam search for solving most relevant explanation in Bayesian networks” by Xiaoyuan Zhu and Changhe Yuan a novel hierarchical beam search algorithm for solving the MRE problem (“Most Relevant Explanation”) in Bayesian networks extends classical beam search algorithms by adding another level for the beam search. The approach is empirically evaluated and shown to outperform classical search algorithms. The paper “An ordered credibility contrast semantics for finite probability agreement” by Paul Snow discusses orderings of subjective probabilities. It proposes a new model grounded in psychological insights that does not rely on interpretations via betting. Liessman Sturlaugson, Logan Perreault, and John W. Sheppard introduce in “Factored performance functions and decision making in continuous time Bayesian networks” a formalization of performance functions for continuous time Bayesian networks. It is shown how these performance functions can be decomposed following the topology of the network and how they can be used for optimization purposes. In “Algebraic model counting” Angelika Kimmig, Guy Van den Broeck, and Luc De Raedt present algebraic model counting as a generalization of weighted model counting, an important problem in many approaches to reasoning with probability. A characterization of algebraic model counting with a specific class of Boolean circuits is established and it is shown how a series of other problems can be represented using the approach developed in the paper. Jean-Christophe Magnan and Pierre-Henri Wuillemin present in “Efficient incremental planning and learning with multi-valued decision diagrams” an approach to use multi-valued decision diagrams as a data structure for planning and learning in discrete domains. The developed algorithms are empirically evaluated and shown to significantly improve previous techniques. Finally, in “What kind of independence do we need for multiple iterated belief change?” the authors Gabriele Kern-Isberner and Daniela Huvermann discuss the problem of multiple iterated belief revision. A novel account for the treatment of independence of information in this process is proposed and related to previous approaches.
Document type :
Directions of work or proceedings
Complete list of metadata
Contributor : Souhila Kaci <>
Submitted on : Thursday, May 24, 2018 - 5:20:01 PM
Last modification on : Friday, April 2, 2021 - 8:38:05 AM

Links full text




Souhila Kaci, Robert Mercer, Matthias Thimm. Editorial. France. Journal of Applied Logic, 22, pp.1-2, 2017, ⟨10.1016/j.jal.2016.11.027⟩. ⟨lirmm-01799445⟩



Record views