On Generators of Random Quasigroup Problems

Problems that can be sampled randomly are a good source of test suites for comparing quality of constraint satisfaction techniques. Quasigroup problems are representatives of structured random problems that are closer to real-life problems and hence more suitable for benchmarking. In this paper, we describe in detail generators for Quasigroup Completion Problem (QCP) and Quasigroups with Holes (QWH). In particular, we study an improvement of the generator for QCP that produces a larger number of satisfiable problems by using propagation through the all-different constraint. We also re-formulate the algorithm for generating QWH that is much faster than the original generator. Finally, we provide an experimental comparison of all presented generators.

• Type: Conference
• Year: 2005
• Where: CSCLP
• Authors: Roman Barták
• Keywords: Constraint Satisfaction Techniques, CSCLP 2005, Problems, Structured Random Problems