In collaboration with: Berthe Choueiry
Meet with Choueiry and me before submitting a proposal on this
project
A local consistency property of a Constraint Satisfaction Problem (CSP) ensures that any subproblem of a given size of the CSP has a solution. Several algorithms for enforcing a given solution property may exist, and the performances of those algorithms typically differ with the input problem. In this project, we will focus on two algorithms for enforcing the minimality property, which was recently shown to be NP-complete. The student will build and characterize a simple classifier that predicts the quickest algorithm for a given CSP instance based on the values of a number of attributes of this instance. A successful solution is a classifier with the smallest error rate while being efficient to execute (CPU time) and having structure transparent to the human user (expressiveness). This project requires a solid background and interest in Constraint Processing and the willingness to learn and use existing code. Significant preliminary work was already carried out, and the potential to publish the results is high.
Return to the CSCE 478/878 Home Page