Explanation-Based Weighted Degree
Refereed Conference Meeting Proceeding
The weighted degree heuristic is among the state of the art generic variable ordering strategies in constraint programming. However, it was often observed that when using large arity constraints, its efficiency deteriorates significantly since it loses its ability to discriminate variables. A possible answer to this drawback is to weight a conflict set rather than the entire scope of a failed constraint. We implemented this method for three common global constraints (AllDifferent, Linear Inequality and Element) and evaluate it on instances from the MiniZinc Challenge. We observe that even with simple explanations, this method outperforms the standard Weighted Degree heuristic.
International Conference on Integration of Artificial Intelligence and Operations Research Techniques in Constraint Programming
Digital Object Identifer (DOI):
National University of Ireland, Cork (UCC)
Open access repository: