Reduced Cost Fixing in MaxSAT

Abstract

We investigate utilizing the integer programming (IP) technique of reduced cost fixing to improve maximum satisfiability (MaxSAT) solving. In particular, we show how reduced cost fixing can be used within the implicit hitting set approach (IHS) for solving MaxSAT. Solvers based on IHS have proved to be quite effective for MaxSAT, especially on problems with a variety of clause weights. The unique feature of IHS solvers is that they utilize both SAT and IP techniques. We show how reduced cost fixing can be used in this framework to conclude that some soft clauses can be left falsified or forced to be satisfied without influencing the optimal cost. Applying these forcings simplifies the remaining problem. We provide an extensive empirical study showing that reduced cost fixing employed in this manner can be useful in improving the state-of-the-art in MaxSAT solving especially on hard instances arising from real-world application domains.