Tuning Randomization in Backtrack Search SAT Algorithms

Abstract

Propositional Satisfiability (SAT) is a well-known NP-complete problem, being fundamental in solving many application problems in Computer Science and Engineering. Recent work on SAT has provided experimental and theoretical evidence that the use of randomization can be quite effective at solving hard instances of SAT. First, randomization was used in local search SAT algorithms, where the search is started over again to avoid getting stuck in a locally optimal partial solution. Moreover, in the last few years randomization has also been included in systematic search algorithms. As a result, backtrack search is given more freedom either to find a solution or to prove unsatisfiability. Indeed, backtrack search algorithms, randomized and run with restarts, were shown to perform significantly better on specific problem instances.