Natural Max-SAT Encoding of Min-SAT

Abstract

We show that there exists a natural encoding which transforms Min-SAT instances into Max-SAT instances. Unlike previous encodings, this natural encoding keeps the same variables, and the optimal assignment for the Min-SAT instance is identical to the optimal assignment of the corresponding Max-SAT instance. In addition to that the encoding can be generalized to the Min-SAT variants with clause weights and hard clauses. We conducted experiments which give evidence that our encoding is practically relevant, as Min-2-SAT instances can be solved much faster by transforming them to Max-SAT and using a Max-SAT solver than by using the best Min-SAT solver directly.