sat求解局部搜索行为分析与概率控制策略word论文.docxVIP

ABSTRACTThe propositional satisfiability problem is the first problem to be proved to be NP-Complete (Non-deterministic Polynomial Complete) by Steven Cook, and there is of great significance in computer science theory and applications. For theory research, it is the cornerstone of the computational complexity theory, and all of the NPC problems can be reduced to SAT problem in polynomial time. For industrial applications, it can be used to solve problems about equivalence checking, bounded model checking, electronic design automation, and statistical physics and so on. Thererfore, efficient and practical algorithms and techniques have been a hot research spot until now.Currently local search algorithms are very effective for solving SAT problem. Sattime is a typical local search algorithm, and has won a silver medal in the random category in the 2011 SAT competition. It is based on Novelty family algorithms and exploits the satisfying history of clauses. The behavior analysis and pattern mining for its solving process uncovered some causal relationships between variables and clauses. We established a database to record local search behavior informations, and excuted data mining on flip event, including association rules mining and frequent sequences mining. Through analyzing the mining results, we found that Sattime sometimes selects the same variable selected in the last step, which may cause loop back, and thus proposed two probability control strategies: strengthen strategy for choosing clauses and strengthen strategy for choosing variables. Applying these two strategies into Sattime algorithm, we designed a new local search algorithm called Sattime-P.We used Sattime and Sattime-P to solve different types of instances, and compared their average flipping numbers and average solving time. Experiment results showed that Sattime-P using improvement strategies has less average flipping numbers and better efficiency than Sattime in most cases.Key words：SAT problem, s