Generalizing Constraint models in Constraint Acquisition
Date:
Our workshop paper on “Generalizing Constraint models in Constraint Acquisition” was presented in the Progress Towards the Holy Grail workshop at the CP Conference.
Abstract: Constraint Acquisition (CA) aims to widen the use of constraint programming by helping users in the modeling process. However, most CA methods learn one ground CSP: a set of individual constraints for a specific problem instance. We focus on generalizing ground CSPs, by learning parameterized constraint models that can model multiple instances of the same problem.
We presented a constraint-level classifier-based approach, where a machine learning classifier is trained to predict for any possible constraint and any possible parameterization of the problem whether the constraint belongs to the problem. A key aspect is an appropriate parameterized feature representation that allows classifiers to learn (implicit) patterns in the ground CSP. The results of our evaluation show that our approach has high accuracy in learning generalized models, and is robust to the presence of noise in the ground CSP.