Open source

Software

Hybrid-Mistral

Hybrid-Mistral is a hybrid CP/SAT solver designed to tackle Disjunctive Scheduling Problems. The SAT engine implements standard Conflict-Driven Clause Learning (CDCL) techniques such as Adaptive Branching (VSIDS), Two-Watched Literals, and 1-UIP Cuts. The latter features are built on top of Mistral-2.0. Propagators are augmented with explanation routines that are invoked only during conflict analysis.

Mistral-2.0

During my PhD, I contributed to the development of the constraint programming solver Mistral-2.0, an open-source C++ library released under the GNU General Public License and available on GitHub. Mistral implements a broad range of arithmetic and logical primitive constraints, as well as several global constraints such as AllDifferent, Linear Equalities, and Element. It also provides standard search strategies and heuristics, including Restarts, Weighted Degree, and Activity-Based Search. The solver won multiple awards in the XCSP constraint programming competition and the minizinc challenge

Tools developed by my students

Faircorels and FairCORELSV2: Python libraries developed mainly by my PhD student Julien Ferry for learning fair and interpretable models. The source code is available here and here
BDDEncoding: A Python library developed by my PhD student Hao Hu for learning binary decision diagrams for classification. The source code is available here .
MaxSAT Decision Trees: A Python library developed by my PhD student Hao Hu for learning optimal decision trees for classification. The source code is available here .