Arc and path consistence revisited
Artificial Intelligence
Generating Satisfiable Problem Instances
Proceedings of the Seventeenth National Conference on Artificial Intelligence and Twelfth Conference on Innovative Applications of Artificial Intelligence
Saving Support-Checks Does Not Always Save Time
Artificial Intelligence Review
Refining the basic constraint propagation algorithm
IJCAI'01 Proceedings of the 17th international joint conference on Artificial intelligence - Volume 1
Probabilistic consistency boosts MAC and SAC
IJCAI'07 Proceedings of the 20th international joint conference on Artifical intelligence
Path consistency by dual consistency
CP'07 Proceedings of the 13th international conference on Principles and practice of constraint programming
Domain consistency with forbidden values
CP'10 Proceedings of the 16th international conference on Principles and practice of constraint programming
Domain consistency with forbidden values
Constraints
| Hi-index | 0.00 |
Arc consistency algorithms are widely used to prune the search space of Constraint Satisfaction Problems (CSPs). Coarse-grained arc consistency algorithms like AC-3, AC-3d and AC-2001 are efficient when it comes to transforming a CSP to its arc-consistent equivalent. These algorithms repeatedly carry out revisions. Revisions require support checks for identifying and deleting all unsupported values from the domain of a variable. In revisions for difficult problems most values have some support. Indeed, most revisions are ineffective, i.e. they cannot delete any value and consume a lot of checks and time. We propose two solutions to overcome these problems. First we introduce the notion of a Support Condition (SC) which guarantees that a value has some support. SCs reduce support checks while maintaining arc consistency during search. Second we introduce the notion of a Revision Condition (RC) which guarantees that all values have support. A RC avoids a candidate revision and queue maintenance overhead. For random problems, SCs reduce the checks required by MAC-3 (MAC-2001) up to 90% (72%). RCs avoid at least 50% of the total revisions. Combining the two results in reducing 50% of the solution time.