Captain Jack: New Variable Selection Heuristics in Local Search for SAT
Stochastic local search (SLS) methods are well known for their ability to find models of randomly generated instances of the propositional satisfiability problem (SAT) very effectively. Two well-known SLS-based SAT solvers are Sparrow, one of the best-performing solvers for random 3-SAT instances, and VE-Sampler, which achieved significant performance improvements over previous SLS solvers on SAT-encoded software verification problems. Here, we introduce a new highly parametric algorithm, Captain Jack, which extends the parameter space of Sparrow to incorporate elements from VE-Sampler and introduces new variable selection heuristics. Captain Jack has a rich design space and can be configured automatically to perform well on various types of SAT instances. We demonstrate that the design space of Captain Jack is easy to interpret and thus facilitates valuable insight into the configurations automatically optimized for different instance sets. We provide evidence that Captain Jack can outperform well-known SLS-based SAT solvers on uniform random k-SAT and ‘industrial-like’ random instances.
KeywordsSearch Step Stochastic Local Search Promising Step Promising Variable Stochastic Local Search Algorithm
Unable to display preview. Download preview PDF.
- 1.Ansótegui, C., Bonet, M.L., Levy, J.: Towards industrial-like random SAT instances. In: IJCAI 2009, pp. 387–392 (2009)Google Scholar
- 2.Balint, A., Diepold, D., Gall, D., Gerber, S., Kapler, G., Retz, R.: EDACC - an advanced platform for the experiment design, administration and analysis of empirical algorithms. In: LION-2011 (to appear)Google Scholar
- 5.bwGRiD: Member of the German D-Grid initiative, funded by the Ministry of Education and Research and the Ministry for Science, Research and Arts Baden-WürttembergGoogle Scholar
- 7.Hoos, H.H.: Computer-aided design of high-performance algorithms. Tech. Rep. TR-2008-16, University of British Columbia (2008)Google Scholar
- 9.Hutter, F., Hoos, H.H., Stützle, T.: Automatic algorithm configuration based on local search. In: AAAI 2007, pp. 1152–1157 (2007)Google Scholar
- 10.KhudaBukhsh, A.R., Xu, L., Hoos, H.H., Leyton-Brown, K.: SATenstein: Automatically building local search SAT solvers from components. In: IJCAI 2009, pp. 517–524 (2009)Google Scholar
- 16.Tompkins, D.A.D.: Dynamic Local Search for SAT: Design, Insights and Analysis. Ph.D. thesis, University of British Columbia (2010)Google Scholar