Abstract
In our work, we investigate the role of redundant clauses in characterizing and solving hard SAT problems. Informally, a redundant clause is one that may be removed from the CNF representation of a SAT instance without altering the satisfying assignments of that instance. Correspondingly, a set of prime clauses is a set of clauses that preserves all the but that contains no redundant clauses. We identify several interesting features of redundant clauses that provide compelling evidence of the correlation between the percentage of redundant clauses and the hardness of instances. We propose a definition of weighted clause-to-variable ratio (WCV), which substantially improves the classic clause-to-variable (m/n) ratio in predicting search cost and explaining the phase transition. WCV is based on a linear combination of the number of prime clauses (NPC) and the number of redundant clauses (NRC). We compare WCV to a number of existing parameters including backbone size and backbone fragility, the constrainedness measure, and the m/n ratio; we posit a variety of advantages to WCV over other measures. We believe that full utilization of redundant knowledge to solve random and real-world SAT problems can significantly improve the performance of SAT solvers, in terms of the scale of the problems that can be dealt with as well as the speed with which these problems are solved.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2005 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Zeng, H., McIlraith, S. (2005). The Role of Redundant Clauses in Solving Satisfiability Problems. In: van Beek, P. (eds) Principles and Practice of Constraint Programming - CP 2005. CP 2005. Lecture Notes in Computer Science, vol 3709. Springer, Berlin, Heidelberg. https://doi.org/10.1007/11564751_111
Download citation
DOI: https://doi.org/10.1007/11564751_111
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-29238-8
Online ISBN: 978-3-540-32050-0
eBook Packages: Computer ScienceComputer Science (R0)