Several local search algorithms for propositional satisfiability have been proposed which can solve hard random problems beyond the range of conventional backtracking procedures. In this paper, we explore the impact of focusing search in these procedures on the "unsatisfied variables"; that is, those variables which appear in clauses which are not yet satisfied. For random problems, we show that such a focus reduces the sensitivity to input parameters. We also observe a simple scaling law in performance. For non-random problems, we show that whilst this focus can improve performance, many problems remain difficult. We speculate that such problems will remain hard for local search unless constraint propagation techniques can be combined with hill-climbing. 1 Introduction Local search is often surprisingly effective as a semi-decision procedure for many NPhard problems. For example, Gsat, a greedy random hill-climbing procedure for propositional satisfiability (or SAT) is very good at ...

A large number of problems that occur in knowledge-representation, learning, VLSI-design, and other areas of artificial intelligence, are essentially satisfiability problems. The satisfiability problem refers to the task of finding a truth assignment that makes a Boolean expression true. The growing need for more efficient and scalable algorithms has led to the development of several SAT solvers. This paper reports the first approach based on combining finite learning automata with metaheuristics. In brief, we introduce a new algorithm that combines finite learning automata with traditional random walk. Furthermore, we present a detailed comparative analysis of the new algorithm's performance, using a benchmark set containing instances from randomized distributions, as well as SAT-encoded problems from various domains.

Abstract. In this paper we compare the performance of three constraint weighting schemes with one of the latest and fastest WSAT heuristics: rnovelty. We extend previous results from satisfiability testing by looking at the broader domain of constraint satisfaction and test for differences in performance using randomly generated problems and problems based on realistic situations and assumptions. We find constraint weighting produces fairly consistent behaviour within problem domains, and is more influenced by the number and interconnectedness of constraints than the realism or randomness of a problem. We conclude that constraint weighting is better suited to smaller structured problems, where it is can clearly distinguish between different constraint groups. 1