Greedy Equivalence Search is a score-based algorithm for learning Bayesian network structure. It works over equivalence classes of directed acyclic graphs, first greedily adding edges that improve the score and then removing edges that no longer help. BDeu, the Bayesian Dirichlet equivalent uniform score, is a common choice for discrete data because it gives the same score to graphs that encode the same independencies. Search cost grows quickly with the number of variables, so systems like ABSOL run GES on many small subsets of columns and aggregate the results.
