Branch-and-Bound Equivalence Search
Branch-and-Bound Equivalence Search is an algorithm for structure learning of Bayesian Networks. It is similar to Greedy Equivalence Search (Chickering, 2002) in that it searches over entire Markov equivalence classes rather than over individual graphs, but it uses the branch-and-bound search strategy to ensure the optimal equivalence class is found.
For details, please see the following paper:
Thijs van Ommen, Learning Bayesian Networks by Branching on Constraints, Proceedings of the Ninth International Conference on Probabilistic Graphical Models (PGM 2018), 2018.
Note that this implementation is a proof-of-concept and has not been optimized.