Home

Declarative Approaches for Enumerating Interesting Patterns



Emmanuel Coquery
Fabien De Marchi
Jean-Marc Petit (Pr) 		Said Jabbour
Lakhdar Sais (Pr)
Karim Tabia 		Marie Agier
Raoul Medina
Lhouari Nourine (Pr)

Objectives

  • Cross-fertilization between artificial intelligence (CSP, SAT), combinatorial algorithmics and databases
  • Bringing original solutions to fundamental pattern mining problems
  • Definition of high level declarative languages for describing interesting pattern enumeration problems

Two major challenges

Challenge 1: Definition of classes of problems for enumerating interesting patterns
Challenge 2: Design of efficient and generic pattern mining systems

Type of datasets

  • Relational databases
  • XML semi-structured data

Examples of patterns

  • Itemsets
  • Sequences
  • Trees
  • Graphs

Interestingness criteria

  • Frequency (or support)
  • Satisfaction of formulas
  • Words
  • Association Rules
  • Functional Dependencies
  • Inclusion Dependencies

Main tasks of the project

  1. Studying the typology of interesting pattern enumeration problems
  2. Conducting theoretical investigations on patterns: structure, languages and interestingness criteria
  3. Applying constraint programming and satisfiability framework for interesting pattern enumeration in large datasets
Job annoucements