Abstract

Stochastic local search algorithms based on the WalkSAT architecture are among the best known methods for solving hard and large instances of the propositional satisfiability problem (SAT). The performance and behaviour of these algorithms critically depends on the setting of the noise parameter, which controls the greediness of the search process. The optimal setting for the noise parameter varies considerably between different types and sizes of problem instances; consequently, considerable manual tuning is typically required to obtain peak performance. In this paper, we characterise the impact of the noise setting on the behaviour of WalkSAT and introduce a simple adaptive noise mechanism for WalkSAT that does not require manual adjustment for different problem instances. We present experimental results indicating that by using this selftuning noise mechanism, various WalkSAT variants (including WalkSAT/SKC and Novelty ) achieve performance levels close to their peak performance for instance-specific, manually tuned noise settings.

Citations

501	Where the really hard problems are - Cheeseman, Kanefsky, et al. - 1991
463	Pushing the envelope: Planning, propositional logic, and stochastic search - KAUTZ, SELMAN - 1996
318	Noise strategies for improving local search - Selman, Kautz, et al. - 1994
192	The reactive tabu search - Battiti, Tecchiolli - 1994
161	Evidence for Invariants in Local Search - McAllester, Selman, et al. - 1997
113	Correlated and uncorrelated fitness landscapes and how to tell the difference - Weinberger - 1990
90	Iterated local search, in - Lourenc-o, Martin, et al. - 2003
77	On the run-time behaviour of stochastic local search algorithms for SAT - Hoos - 1999
58	When gravity fails: Local search topology - Frank, Cheeseman, et al. - 1997
57	SATLIB: An online resource for research on SAT - Hoos, Stützle - 2000
56	T.: Local search algorithms for SAT: An empirical evaluation - Hoos, Stutzle
53	Local search characteristics of incomplete SAT procedures - Schuurmans, Southey - 2000
39	Tuning local search for satisfiability testing - Parkes, Walser - 1996
39	The exponentiated subgradient algorithm for heuristic boolean programming - Schuurmans, Southey, et al. - 2001
38	Towards a characterisation of the behaviour of stochastic local search algorithms for SAT - Hoos, Stützle - 1999
2	A Self-Tuning Implementation of Walksat - Patterson, Kautz - 2001