A Parametric Approach for Smaller and Better Encodings of Cardinality Constraints

Abstract

Abstract. Adequate encodings for high-level constraints are a key ingredient for the application of SAT technology. In particular, cardinality constraints state that at most (at least, or exactly) k out of n propositional variables can be true. They are crucial in many applications. Although sophisticated encodings for cardinality constraints exist, it is well known that for small n and k straightforward encodings without auxiliary variables sometimes behave better, and that the choice of the right trade-off between minimizing either the number of variables or the number of clauses is highly applicationdependent. Here we build upon previous work on Cardinality Networks to get the best of several worlds: we develop an arc-consistent encoding that, by recursively decomposing the constraint into smaller ones, allows one to decide which encoding to apply to each sub-constraint. This process minimizes a function λ · num vars + num clauses, being λ a parameter that can be tuned by the user. Our careful experimental evaluation shows that (e.g., for λ = 5) this new technique produces much smaller encodings in variables and clauses, and indeed strongly improves SAT solvers' performance.

Cost-optimal constrained correlation clustering via weighted partial Maximum Satisfiability

“…x iN to be set to true, which can further be encoded with hard clauses using one of the several compact cardinality constraints; see e.g. [57,58]. The whole AtMostInAll(k) constraint decomposes in to a conjunction of such cardinality constraints over i.…”

Quantitative Verification of Neural Networks and Its Security Applications

“…Observe that we can express each block in BNN as a conjunction of cardinality constraints [4,8,91]. Cardinality constraints are constraints over boolean variables x 1 , .…”

“…We choose cardinality networks [4,8] to encode the cardinality constraints to CNF formulas and show for this particular encoding that the resulting CNF is equi-witnessable to the cardinality constraint. Cardinality networks implement several types of gates, i.e., merge circuits, sorting circuits and 2-comparators, that compose to implement a merge sort algorithm.…”

νZ - An Optimizing SMT Solver

“…νZ also contains a Pseudo-Boolean theory solver. It borrows from [4,1] for simplification, generating conflict clauses, and incrementally compiling into small sorting circuits. It also adds an option to prune branches using dual simplex.…”