We present and evaluate a new compiler, called d4, targeting the Decision-DNNF language. As the state-of-the-art compilers C2D and Dsharp targeting the same language, d4 is a top-down tree-search algorithm exploring the space of propositional interpretations. d4 is based on the same ingredients as those considered in C2D and Dsharp (mainly, disjoint component analysis, conflict analysis and non-chronological backtracking, component caching). d4 takes advantage of a dynamic decomposition approach based on hypergraph partitioning, used sparingly. Some simplification rules are also used to minimize the time spent in the partitioning steps and to promote the quality of the decompositions. Experiments show that the compilation times and the sizes of the Decision-DNNF representations computed by d4 are in many cases significantly lower than the ones obtained by C2D and Dsharp.

No abstract

In this paper, we investigate the computational intelligibility of Boolean classifiers, characterized by their ability to answer XAI queries in polynomial time. The classifiers under consideration are decision trees, DNF formulae, decision lists, decision rules, tree ensembles, and Boolean neural nets. Using 9 XAI queries, including both explanation queries and verification queries, we show the existence of large intelligibility gap between the families of classifiers. On the one hand, all the 9 XAI queries are tractable for decision trees. On the other hand, none of them is tractable for DNF formulae, decision lists, random forests, boosted decision trees, Boolean multilayer perceptrons, and binarized neural networks.

No abstract

No abstract

The enumeration of all Maximal Satisfiable Subsets (MSSes) or all Minimal Correction Subsets (MCSes) of an unsatisfiable CNF Boolean formula is a useful and sometimes necessary step for solving a variety of important A.I. issues. Although the number of different MCSes of a CNF Boolean formula is exponential in the worst case, it remains low in many practical situations; this makes the tentative enumeration possibly successful in these latter cases. In the paper, a technique is introduced that boosts the currently most efficient practical approaches to enumerate MCSes. It implements a model rotation paradigm that allows the set of MCSes to be computed in an heuristically efficient way.

No abstract

Abstract. In earlier work on a limited form of extended resolution for CDCL based SAT solving, new literals were introduced to factor out parts of learned clauses. The main goal was to shorten clauses, reduce proof size and memory usage and thus speed up propagation and conflict analysis. Even though some reduction was achieved, the effectiveness of this technique was rather modest for generic SAT solving. In this paper we show that factoring out literals is particularly useful for incremental SAT solving, based on assumptions. This is the most common approach for incremental SAT solving and was pioneered by the authors of MINISAT. Our first contribution is to focus on factoring out only assumptions, and actually all eagerly. This enables the use of compact dedicated data structures, and naturally suggests a new form of clause minimization, our second contribution. As last main contribution, we propose to use these data structures to maintain a partial proof trace for learned clauses with assumptions, which gives us a cheap way to flush useless learned clauses. In order to evaluate the effectiveness of our techniques we implemented them within the version of MINISAT used in the publically available state-of-the-art MUS extractor MUSer. An extensive experimental evaluation shows that factoring out assumptions in combination with our novel clause minimization procedure and eager clause removal is particularly effective in reducing average clause size, improves running time and in general the state-of-the-art in MUS extraction.