• HDR Defended on :
  • Dec 6, 2018 • 14h, salle des thèses de la faculté Jean Perrin

Abstract

My research work is at the boundary of several fields including symbolic Artificial Intelligence (AI), data mining, graphs and information systems. It is divided into the following keywords: constrained reasoning, knowledge representation, reasoning modeling, constrained pattern mining, clustering, community detection, data compression and web services composition.

The first part of my HDR gives an overview of my algorithmic contributions to the problem of propositional satisfiability including clause learning, parallel portfolio SAT solvers, enumeration of models and prime implicants, and transformations of cardinality constraints in conjunctive normal form.

The second part deals with data mining and clustering. My contributions focus on declarative approaches for different data mining tasks: frequent itemsets mining and their various condensed forms, association rule mining and its many variants, sequential pattern mining, Top-k pattern enumeration modulo a preference relation, pattern mining under uncertainty, parallel decomposition methods and symbolic clustering of propositional formulas. To highlight the cross-fertilizations between symbolic AI and data mining, I show on the one hand how the concept of symmetry widely explored in SAT/CP is extended to the mining of set patterns, and on the other hand, how data mining can be exploited to compress Boolean formulas and CSP constraints expressed in extension.

The third part deals with my contributions to community detection and compression of large graphs, exploiting pseudo-Boolean constraints and propositional logic.

The fourth part focuses on reasoning in the presence of inconsistencies and argumentation theory. I present different methods for the quantification of conflicts in knowledge bases and also for reasoning in the presence of inconsistencies and uncertainty in ontologies.

The last part is devoted to my contributions to the composition of web services, by proposing various logic models of this problem.

This work ends by mentioning my current and future research projects.

Jury

  • Sihem Amer-Yahya, Research Director, LIG - CNRS, University of Grenoble Alpes, France (Rapporteur)
  • Philippe Besnard, Research Director, IRIT - CNRS, Paul Sabatier University, France
  • Philippe Fournier-Viger, Professor, Harbin Institute of Technology, China (Rapporteur)
  • Anthony Hunter, Professor, University College London, United Kingdom
  • Felip Manyà, Professor, IIIA-CSIC, Spain (Rapporteur)
  • Jean-Charles Régin, Professor, I3S, University of Nice-Sophia Antipolis, France
  • Lakhdar Saïs, University Professor, University of Artois, France