Unweighted Stochastic Local Search can be Effective for Random CSP Benchmarks

Christopher D. Rosin

September, 2014

ULSA is a simple unweighted stochastic local searcher for random constraint satisfaction problems (CSP). It achieves record results on the BHOSLIB benchmarks (converted to CSP instances), including the first reported score of 99 on the challenging frb100-40 instance. ULSA is described in:

• Rosin, C.D. (2014). Unweighted stochastic local search can be effective for random CSP benchmarks. arXiv:1411.7480. PDF.

Download source code for the ULSA stochastic local search solver for random binary CSPs.