SATMiner

SATMiner is a library for searching interesting patterns using constraint programming/SAT backend. It was initially developed as a prototype for the DAG project, in collaboration with people at CRIL.

It is currently used as a basis for experiments on SAT/pseudo boolean based pattern mining on:

• sequences
• attributes sets in relational data (SatQL language)

The current version is coded in Java and uses SAT4J and a modified version of MiniSat as backends. It also uses the BoolVar/PB library, as well as an implementation of cardinality networks for translation of some pseudo-boolean constraints to SAT.

GIT, stable branch: https://liris.cnrs.fr/~ecoquery/satminer.git, lastest stable binary jar file

• source (2014-05-12, changed syntax for SatQL¹⁾: now SQL expressions (queries and conditions) should be in “{ }” braces and Java joins are not available anymore. However arbitrary SQL expressions can now be used)
• source (2014-04-15, added IF THEN, LIKE, LIMIT to SatQL, version including bench datasets)
• source (2014-02-18, includes Boolvar/PB and a modified version of minisat, fixes compilation problems)
• source (2014-02-17, includes Boolvar/PB and a modified version of minisat)

• Emmanuel Coquery (LIRIS)
• Said Jabbour (CRIL)
• Jean-Marc Petit (LIRIS)
• Lakhdar Sais (CRIL)

• Emmanuel Coquery, Saïd Jabbour, Lakhdar Saïs, Yakoub Salhi: A SAT-Based Approach for Discovering Frequent, Closed and Maximal Patterns in a Sequence. ECAI 2012: 258-263 pdf
• Emmanuel Coquery, Jean-Marc Petit, Lakhdar Sais: Using SAT and SQL for Pattern Mining in relational databases. First workshop on COmbining COnstraint solving with MIning and LEarning, 2012. pdf
• Emmanuel Coquery, Saïd Jabbour, Lakhdar Sais: A Constraint Programming Approach for Enumerating Motifs in a Sequence. ICDM Workshops 2011, WDPM: 1091-1097 pdf