Ignasi Abío

A new look at BDDs for Pseudo-Boolean constraints

Pseudo-Boolean constraints are omnipresent in practical applications, and thus a significant effort has been devoted to the development of good SAT encoding techniques for them. Some of these encodings first construct a Binary Decision Diagram (BDD) for the constraint, and then encode the BDD into a propositional formula. These BDD-based approaches have some important advantages, such as not being dependent on the size of the coefficients, or being able to share the same BDD for representing many constraints. We first focus on the size of the resulting BDDs, which was considered to be an open problem in our research community. We report on previous work where it was proved that there are Pseudo-Boolean constraints for which no polynomial BDD exists. We also give an alternative and simpler proof assuming that NP is different from Co-NP. More interestingly, here we also show how to overcome the possible exponential blowup of BDDs by coefficient decomposition. This allows us to give the first polynomial generalized arc-consistent ROBDD-based encoding for Pseudo-Boolean constraints. Finally, we focus on practical issues: we show how to efficiently construct such ROBDDs, how to encode them into SAT with only 2 clauses per node, and present experimental results that confirm that our approach is competitive with other encodings and state-of-the-art Pseudo-Boolean solvers.

Conflict directed lazy decomposition

Two competing approaches to handling complex constraints in satisfaction and optimization problems using SAT and LCG/SMT technology are: decompose the complex constraint into a set of clauses; or (theory) propagate the complex constraint using a standalone algorithm and explain the propagation. Each approach has its benefits. The decomposition approach is prone to an explosion in size to represent the problem, while the propagation approach may require exponentially more search since it does not have access to intermediate literals for explanation. In this paper we show how we can obtain the best of both worlds by lazily decomposing a complex constraint propagator using conflicts to direct it. If intermediate literals are not helpful for conflicts then it will act like the propagation approach, but if they are helpful it will act like the decomposition approach. Experimental results show that it is never much worse than the better of the decomposition and propagation approaches, and sometimes better than both.

BDDs for pseudo-boolean constraints: revisited

Pseudo-Boolean constraints are omnipresent in practical applications, and therefore a significant effort has been devoted to the development of good SAT encoding techniques for these constraints. Several of these encodings are based on building Binary Decision Diagrams (BDDs) and translating these into CNF. Indeed, BDD-based encodings have important advantages, such as sharing the same BDD for representing many constraints. Here we first prove that, unless NP = Co-NP, there are Pseudo-Boolean constraints that admit no variable ordering giving a polynomial (Reduced, Ordered) BDD. As far as we know, this result is new (in spite of some misleading information in the literature). It gives several interesting insights, also relating proof complexity and BDDs. But, more interestingly for practice, here we also show how to overcome this theoretical limitation by coefficient decomposition. This allows us to give the first polynomial arc-consistent BDD-based encoding for Pseudo-Boolean constraints.

Encoding Linear Constraints into SAT

Linear integer constraints are one of the most important constraints in combinatorial problems since they are commonly found in many practical applications. Typically, encoding linear constraints to SAT performs poorly in problems with these constraints in comparison to constraint programming (CP) or mixed integer programming (MIP) solvers. But some problems contain a mix of combinatoric constraints and linear constraints, where encoding to SAT is highly effective. In this paper we define new approaches to encoding linear constraints into SAT, by extending encoding methods for pseudo-Boolean constraints. Experimental results show that these methods are not only better than the state-of-the-art SAT encodings, but also improve on MIP and CP solvers on appropriate problems. Combining the new encoding with lazy decomposition, which during runtime only encodes constraints that are important to the solving process that occurs, gives a robust approach to many highly combinatorial problems involving linear constraints.

The ultrametric space of plane branches

We study properties of the space of irreducible germs of plane curves (branches), seen as an ultrametric space. We provide various geometrical methods to measure the distance between two branches and to compare distances between branches, in terms of topological invariants of the singularity which comprises some of the branches. We show that, in spite of being very close to the notion of intersection multiplicity between two germs, this notion of distance behaves very differently. For instance, any value in [0, 1] ∩ ℚ is attained as the distance between a fixed branch and some other branch, in contrast with the fact that the semigroup of the fixed branch has gaps. We also present results that lead to interpret this distance as a sort of geometric distance between the topological equivalence or equisingularity classes of branches.

Encoding Linear Constraints with Implication Chains to CNF

Linear constraints are the most common constraints occurring in combinatorial problems. For some problems which combine linear constraints with highly combinatorial constraints, the best solving method is translation to SAT. Translation of a single linear constraint to SAT is a well studied problem, particularly for cardinality and pseudo-Boolean constraints. In this paper we describe how we can improve encodings of linear constraints by taking into account implication chains in the problem. The resulting encodings are smaller and can propagate more strongly than separate encodings. We illustrate benchmarks where the encoding improves performance.

On CNF Encodings of Decision Diagrams

Decisions diagrams such as Binary Decision Diagrams (BDDs), Multi-valued Decision Diagrams (MDDs) and Negation Normal Forms (NNFs) provide succinct ways of representing Boolean and other finite functions. Hence they provide a powerful tool for modelling complex constraints in discrete satisfaction and optimization problems. Generic propagators for these global constraints exist, but they are complex and hard to implement. An alternative approach to making use of them for solving is to encode them to CNF, using SAT style solving technology to implement them efficiently. This may also have advantages since it is naturally incremental and exposes intermediate literals which may well be useful as search decisions for solving the problem.