We examine the behavior of dynamic value-ordering heuristics in a CSP under the requirement to generate a large number of diverse solutions as fast as possible. In particular, we analyze the trade-off between the solution search performance and the diversity of the generated solutions, and propose a general probabilistic approach to control and improve this trade-off. Several old/new learning-reuse heuristics are described, extending the survivors-first value-ordering heuristics family. The proposed approach is illustrated on a real-world set of examples from the Automatic Test Generation problem domain, as well as on several sets of random binary CSPs. © 2010 Springer-Verlag.
CITATION STYLE
Schreiber, Y. (2010). Value-ordering heuristics: Search performance vs. solution diversity. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 6308 LNCS, pp. 429–444). Springer Verlag. https://doi.org/10.1007/978-3-642-15396-9_35
Mendeley helps you to discover research relevant for your work.