Gilles Audemard

Gilles Audemard
Université d'Artois · CRIL Centre de Recherche en Informatique de Lens

About

87
Publications
7,625
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
1,897
Citations
Citations since 2017
16 Research Items
913 Citations
2017201820192020202120222023050100150200
2017201820192020202120222023050100150200
2017201820192020202120222023050100150200
2017201820192020202120222023050100150200

Publications

Publications (87)
Article
Decision trees have long been recognized as models of choice in sensitive applications where interpretability is of paramount importance. In this paper, we examine the computational ability of Boolean decision trees for the explanation purpose. We focus on both abductive explanations (suited to explaining why a given instance has been classified as...
Preprint
Full-text available
Boosted trees is a dominant ML model, exhibiting high accuracy. However, boosted trees are hardly intelligible, and this is a problem whenever they are used in safety-critical applications. Indeed, in such a context, rigorous explanations of the predictions made are expected. Recent work have shown how subset-minimal abductive explanations can be d...
Preprint
Full-text available
This document represents the proceedings of the 2022 XCSP3 Competition. The results of this competition of constraint solvers were presented at FLOC (Federated Logic Conference) 2022 Olympic Games, held in Haifa, Israel from 31th July 2022 to 7th August, 2022.
Conference Paper
Abductive explanations take a central place in eXplainable Artificial Intelligence (XAI) by clarifying with few features the way data instances are classified. However, instances may have exponentially many minimum-size abductive explanations, and this source of complexity holds even for ``intelligible'' classifiers, such as decision trees. When th...
Article
Random forests have long been considered as powerful model ensembles in machine learning. By training multiple decision trees, whose diversity is fostered through data and feature subsampling, the resulting random forest can lead to more stable and reliable predictions than a single decision tree. This however comes at the cost of decreased interpr...
Conference Paper
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 explanatio...
Preprint
Full-text available
Decision trees have long been recognized as models of choice in sensitive applications where interpretability is of paramount importance. In this paper, we examine the computational ability of Boolean decision trees in deriving, minimizing, and counting sufficient reasons and contrastive explanations. We prove that the set of all sufficient reasons...
Preprint
Full-text available
Random forests have long been considered as powerful model ensembles in machine learning. By training multiple decision trees, whose diversity is fostered through data and feature subsampling, the resulting random forest can lead to more stable and reliable predictions than a single decision tree. This however comes at the cost of decreased interpr...
Preprint
Full-text available
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 explanatio...
Conference Paper
Full-text available
SAT research has a long history of source code and binary releases, thanks to competitions organized every year. However, since every cycle of competitions has its own set of rules and an adhoc way of publishing source code and binaries, compiling or even running any solver may be harder than what it seems. Moreover, there has been more than a thou...
Preprint
Full-text available
SAT research has a long history of source code and binary releases, thanks to competitions organized every year. However, since every cycle of competitions has its own set of rules and an adhoc way of publishing source code and binaries, compiling or even running any solver may be harder than what it seems. Moreover, there has been more than a thou...
Conference Paper
Full-text available
With the advent of multi-core processors, it makes sense to design multithreaded solvers. Nevertheless, implementing such solvers is often a cumbersome task. Indeed, multithreaded SAT solvers are not easy to write, and only experienced programmers should undertake to code for these types of applications. To overcome this problem, we propose a new l...
Article
The set of novelties introduced with the SAT solver Glucose is now considered as a standard for practical SAT solving. In this paper, we review the different strategies and technologies added in Glucose over the years. We detail each technique and discuss its impact on the final performances reached by Glucose. We also come back on one of the main...
Conference Paper
A portfolio SAT solver has to share clauses in order to be efficient. In a distributed environment, such sharing implies additional problems: more information has to be exchanged and communications among solvers can be time consuming. In this paper, we propose a new version of the state-of-the-art SAT solver Syrup that is now able to run on distrib...
Conference Paper
We present and evaluate AmPharoS, a new parallel SAT solver based on the divide and conquer paradigm. This solver, designed to work on a great number of cores, runs workers on sub-formulas restricted to cubes. In addition to classical clause sharing, it also exchange extra information associated to the cubes. Furthermore, we propose a new criterion...
Conference Paper
With the increasing performance of SAT solvers, a lot of distinct problems, coming from very disparate fields, are added to the pool of Application problems, regularly used to rank solvers. These problems are also widely used to measure the positive impact of any new idea. We show in this paper that many of them have extreme behaviors that any SAT...
Conference Paper
Full-text available
Over the years, parallel SAT solving becomes more and more important. However, most of state-of-the-art parallel SAT solvers are portfolio-based ones. They aim at running several times the same solver with different parameters. In this paper, we propose a tool called Dolius, mainly based on the divide and conquer paradigm. In contrast to most curre...
Conference Paper
Full-text available
Modern CDCL SAT solvers routinely solve very large industrial SAT instances in relatively short periods of time. It is clear that these solvers somehow exploit the structure of real-world instances. However, to-date there have been few results that precisely characterise this structure. In this paper, we provide evidence that the community structur...
Conference Paper
Managing learnt clauses among a parallel, memory shared, SAT solver is a crucial but difficult task. Based on some statistical experiments made on learnt clauses, we propose a simple parallel version of Glucose that uses a lazy policy to exchange clauses between cores. This policy does not send a clause when it is learnt, but later, when it has a c...
Article
Full-text available
Most of state-of-the-art parallel SAT solvers are portfolio-based ones. They aim at running several times the same solver with different parameters. In this paper, we propose a solver called Dolius, based on the divide and conquer paradigm. In contrast to most current parallel efficient engines, Dolius does not need shared memory, can be distributed...
Technical Report
Full-text available
This paper provides a short system description of our updated portfolio-based solver called PeneLoPe, based on ManySat. Particularly, this solver focuses on collaboration between threads, providing different policies for exporting and importing learnt clauses between CDCL searches. Moreover, different restart strategies are also available, together...
Article
Full-text available
This note describes features of the version of SATHYS that entered the SAT-race 2010 affiliated to the SAT'2010 conference in Edinburgh, Scotland, UK. 1 Overview SATHYS (Sat Hybrid Solver) is a new hybridization of local search and modern SAT solver. In our approach, both components heavily cooperate through relevant informa-tion gathered during se...
Article
Full-text available
Proceedings of SAT Race 2010: Solver and Benchmarks Descriptions
Conference Paper
Since the first principles of Knowledge Compilation (KC), most of the work has been focused in finding a good compilation target language in terms of compromises between compactness and expressiveness. The central idea remains unchanged in the last fifteen years: an off-line, very hard, stage, allows to "compile" the initial theory in order to guar...
Chapter
Full-text available
This paper provides a short system description of our updated portfolio-based solver called PeneLoPe, based on ManySat. Particularly, this solver focuses on collaboration between threads, providing different policies for exporting and importing learnt clauses between CDCL searches. Moreover, different restart strategies are also available, together...
Conference Paper
Beside the important progresses observed in SAT solving, a number of applications explicitly rely on incremental SAT solving only. In this paper, we focus on refining the incremental SAT Solver Glucose, from the SAT engine perspective, and address a number of unseen problems this new use of SAT solvers opened. By playing on clause database cleaning...
Article
In this paper, we propose a new distributed approach for SAT. This approach, based on the divide-andconquer principles, relies on master/slave architecture, where each slave is a CDCL solver working on a subformula of the initial problem, whereas the master does not actively participate to the solving process, since its main task is to coordonate t...
Conference Paper
So-called Modern SAT solvers are built upon a few – but essential – ingredients: branching, learning, restarting and clause database cleaning. Most of them have been greatly improved since their first introduction, more than ten years ago. In many cases, the initial reasons that lead to their introduction do not explain anymore their current usage...
Conference Paper
Full-text available
Managing learnt clause database is known to be a tricky task in SAT solvers. In the portfolio framework, the collaboration between threads through learnt clause exchange makes this problem even more difficult to tackle. Several techniques have been proposed in the last few years, but practical results are still in favor of very limited collaboratio...
Conference Paper
Full-text available
In this paper, we propose a new dynamic management policy of the learnt clause database in modern SAT solvers. It is based on a dynamic freezing and activation principle of the learnt clauses. At a given search state, using a relevant selection function, it activates the most promising learnt clauses while freezing irrelevant ones. In this way, cla...
Conference Paper
Full-text available
In this paper, a novel hybrid and complete approach for propositional satisfiability, called SatHys (Sat Hybrid Solver), is introduced. It efficiently combines the strength of both local search and cdcl based sat solvers. Considering the consistent partial assignment under construction by the cdcl sat solver, local search is used to extend it to a...
Conference Paper
Modern complete SAT solvers almost uniformly implement variations of the clause learning framework introduced by Grasp and Chaff. The success of these solvers has been theoretically explained by showing that the clause learning framework is an implementation of a proof system which is as poweful as resolution. However, exponential lower bounds are...
Article
Modern complete SAT solvers almost uniformly implement variations of the clause learning framework introduced by Grasp and Chaff. The success of these solvers has been theoretically explained by showing that the clause learning framework is an implementation of a proof system which is as poweful as resolution. However, exponential lower bounds are...
Article
La plupart des solveurs SAT modernes se basent, avec succès, sur les mécanismes d'analyse de conflits et d'apprentissage initialement introduits dans les solveurs GRASP et CHAFF. D'un point de vue théorique, ce succès a été partiellement expliqué à l'aide de l'équivalence, en termes de puissance, entre le système de preuves implanté par l'apprentis...
Article
Full-text available
Cet article introduit SATHYS (SAT Hybrid Solver), une nouvelle approche hybride pour le problème de la satisfia-bilité propositionnelle (SAT). Cette approche combine la recherche locale et un solveur CDCL (« Conflict Driven Clause Learning »). Chaque fois qu'un minimum local est atteint par la recherche locale, un ensemble de variables est fixé à l...
Conference Paper
Full-text available
In this paper a learning based local search approach for propositional satisfiability is presented. It is based on an original adaptation of the conflict driven clause learning (CDCL) scheme to local search. First an extended implication graph for complete assignments of the set of variables is proposed. Secondly, a unit propagation based technique...
Conference Paper
Full-text available
This article introduces SatHyS (SAT HYbrid Solver), a novel hybrid approach for propositional satisfiability. It combines local search and conflict driven clause learning (CDCL) scheme. Each time the local search part reaches a local minimum, the CDCL is launched. For SAT problems it behaves like a tabu list, whereas for UNSAT ones, the CDCL part...
Conference Paper
Beside impressive progresses made by SAT solvers over the last ten years, only few works tried to un- derstand why Conflict Directed Clause Learning algorithms (CDCL) are so strong and efficient on most industrial applications. We report in this work a key observation of CDCL solvers behavior on this family of benchmarks and explain it by an unsus-...
Article
Full-text available
Dans cet article, nous présentons une nouvelle approche pour sortir des minimums locaux dans le cadre de la recherche locale. Cette approche est basée sur le principe d'analyse de conflits utilisé dans les solveurs SAT modernes. Nous proposons une extension du graphe d'implications au cadre de la recherche locale où plusieurs conflits sont présents...
Article
GLUCOSE is based on a new scoring scheme for the clause learning mechanism, based on the paper (Audemard and Simon, 2009). This short competi- tion report summarizes the techniques embedded in the competition 09 version of GLUCOSE. Solver's name comes from glue clauses, a particular kind of clauses that GLUCOSE detects and preserves during search....
Conference Paper
Experimentation of new algorithms is the usual companion section of papers dealing with SAT. However, the behavior of those algorithms is so unpredictable that even strong experiments (hundreds of benchmarks, dozen of solvers) can be still misleading. We present here a set of experiments of very small changes of a canonical Conflict Driven Clause L...
Article
In this paper a new graph based representation of Boolean formulas in conjunctive normal form (CNF) is proposed. It extends the well-known graph representation of binary CNF formulas (2-SAT) to the general case. Every clause is represented as a set of (conditional) implications and encoded with different edges labeled with a set of literals, called...
Article
Full-text available
Cet article présente plusieurs contributions au " Conflict Driven Clauses Learning " (CDCL), qui est une des composantes clés des solveurs SAT modernes. Tout d’abord, nous montrons que, à partir du graphe d’implication, les clauses assertives obtenues en utilisant le principe du premier point d’implication unique (" First Unique Implication Point "...
Conference Paper
Full-text available
This paper makes several contributions to Conflict Driven Clauses Learning (CDCL), which is one of the key components of modern SAT solvers. First, we show that, given an implication graph, asserting clauses derived us- ing the first Unique Implication Point principle (First UIP) are optimal in terms of back-jumping. Secondly we propose an extended...
Conference Paper
In this paper a new circuit sat based encoding of boolean formula is proposed. It makes an original use of the concept of restrictive models introduced by Boufkhad to polynomially translate any formula in conjunctive normal form (CNF) to a circuit sat representation (a conjunction of gates and clauses). Our proposed encoding preserves the satisfiab...
Conference Paper
Local search algorithms for satisfiability testing are still the best methods for a large number of prob- lems, despite tremendous progresses observed on complete search algorithms over the last few years. However, their intrinsic limit does not allow them to address UNSAT problems. Ten years ago, this question challenged the community without any...
Conference Paper
Full-text available
Many reasoning task and combinatorial problems exhibit symmetries. Exploiting such symmetries has been proved to be very important in reducing search efforts. Breaking symmetries using additional constraints is currently one of the most used approaches. Extending such symmetry breaking techniques to quantified boolean formulae (QBF) is a very chall...
Article
Full-text available
Symmetries abound in logically formulated problems where many axioms are universally quantified, as this is the case in equational theories. Two complementary approaches have been used so far to dynamically tackle those symmetries: prediction and detection. The best-known predictive symmetry elimination method is the least number heuristic (lnh). A...
Article
De nombreuses tâches et problèmes combinatoires exhibent des symétries. La résolution de tels problèmes conduit à répéter inlassablement l'étude de situations ou de sous- problèmes équivalents. Depuis plusieurs années, l'exploitation des symétries a permis une réduction significative de l'espace de recherche et la résolution de problèmes ouverts ju...
Conference Paper
Solving Quantified Boolean Formulas (QBF) has become an impor- tant and attractive research area, since several problem classes might be formu- lated efficiently as QBF instances (e.g. planning, non monotonic reasoning, two- player games, model checking, etc). Many QBF solvers has been proposed, most of them perform decision tree search using the D...
Article
Résumé Depuis quelques années, la résolution des for-mules booléennes quantifiées (QBF) est devenu un domaine de recherche important et attractif. En effet, de nombreuses classes de problèmes peuvent être formulées efficacement en des instances QBF (pla-nification, raisonnement non monotone, model che-cking). Beaucoup de solveurs ont été proposé. L...
Article
Full-text available
Industrial systems of practical relevance can be often characterized in terms of discrete control variables and real-valued physical variables, and can therefore be modeled as hybrid automata. Unfortunately, continuity of the physical behaviour over time, or triangular constraints, must often be assumed, which yield an undecidable class of hybrid a...
Conference Paper
Solving quantified Boolean formulas (QBF) has become an attractive research area in artificial intelligence. Many important artificial intelligence problems (planning, nonmonotonic reasoning, formal verification, etc.) can be reduced to QBFs. A new DLL-based method is proposed that integrates binary decision diagram (BDD) to set free the variable o...
Conference Paper
Full-text available
Many reasoning task and combinatorial problems exhibit symmetries. Exploiting symmetries has been proved very important in reducing search efforts. This important task is widely investigated in constraint satisfaction problems and satisfiability of boolean formulas. In this paper, we show how symmetries can be naturally extended to Quantified Boole...
Article
Full-text available
This paper presents the OpenSAT project, an open source framework of SAT algorithms in Java. The objective of this framework is twofold: first, to provide some specifications about SAT solver services for a better interoperability between SAT solvers and SAT powered applications, and second, to give a reference implementation of different specifica...
Article
In the last years we have witnessed an impressive advance in the eciency of boolean solving techniques, which has brought large previously intractable problems at the reach of state-of-the-art solvers.
Conference Paper
Full-text available
Enormous progress has been achieved in the last decade in the verification of timed systems, making it possible to analyze significant real-world protocols. An open challenge is the identification of fully symbolic verification techniques, able to deal effectively with the finite state component as well as with the timing aspects. In this paper we...
Article
Enormous progress has been achieved in the last decade in the veri cation of timed systems, making it possible to verify signi cant real-world protocols. An open challenge is the identi cation of fully symbolic veri cation techniques, able to deal eectively with the nite state component as well as with the timing aspects.
Conference Paper
Finite model search for first-order logic theories is complementary to theorem proving. Systems like Falcon, SEM and FMSET use the known LNH (Least Number Heuristic) heuristic to eliminate some trivial symmetries. Such symmetries are worthy, but their exploitation is limited to the first levels of the model search tree, since they disappear as soon...
Conference Paper
The availability of decision procedures for combinations of boolean and linear mathematical propositions opens the ability to solve problems arising from real-world domains such as verification of timed systems and planning with resources. In this paper we present a general and efficient approach to the problem, based on two main ingredients. The f...
Article
Full-text available
The availability of decision procedures for combinations of boolean and linear mathematical propositions opens the ability to solve problems arising from real-world domains such as verification of timed systems and planning with resources. In this paper we present a general and efficient approach to the problem, based on two main ingredients. The f...
Article
Full-text available
Enormous progress has been achieved in the last decade in the verification of timed systems, making it possible to verify significant real-world protocols. An open challenge is the identification of fully symbolic verification techniques, able to deal effectively with the finite state component as well as with the timing aspects.
Chapter
Full-text available
In the last years we have witnessed an impressive advance in the efficiency of boolean solving techniques, which has brought large previously intractable problems at the reach of state-of-the-art solvers. Unfortunately, simple boolean expressions are not expressive enough for representing many real-world problems, which require handling also intege...
Conference Paper
Full-text available
This paper presents an algorithm, XLNH, to generate finite models of first order equational theories. Unlike conventional methods, which focus on using as few individual constants as possible to preserve symmetries, XLNH heuristically selects then fully generates the functions that appear in the problem, using a weighted directed graph of functiona...
Article
Systems like Falcon, SEM and FMSET use the known LNH (Least Number Heuristic) heuristic to eliminate some trivial symmetries.
Conference Paper
. This article introduces two techniques to improve the propagation eciency of CSP based nite model generation methods. One approach consists in statically rewriting some selected clauses so as to trigger added constraint propagations. The other approach uses a dynamic lookahead strategy to both lter out inconsistent domain values and select the mo...
Article
Full-text available
. We study an algorithm for the SAT problem which is based on the Davis & Putnam procedure. The main idea is to increase the application of the unit clause rule during the search. When there is no unit clause in the set of clauses, our method tries to produce one occuring in the current subset of binary clauses. A literal deduction algorithm is imp...
Conference Paper
Finite modele search for first order logic theories is a complementary alternative to automated deduction. Systems like Falcon, SEM and FMSET use the LNH(Least Number Heuristic) heuristic to eliminate some trivial symmetries. Such symmetries are worthful, but their exploitation is limited to the first levels of the model search tree, since they des...
Conference Paper
Full-text available
We study an algorithm for the SAT problem which is based on the Davis & Putnam procedure. The main idea is to increase the ap- plication of the unit clause rule during the search. When there is no unit clause in the set of clauses, our method tries to produce one occuring in the current subset of binary clauses. A literal deduction algorithm is imp...
Article
Full-text available
Résumé Nous prsentons un algorithme permettant de gnrer des modles finis de thories du premier ordre. Cet algorithme s'avre trs performant sur une catgorie particulire des thories quationnelles, les groupes abliens, qui rsistent aux approches connues. Pen-dant la recherche, et sans surcot de temps, la plupart des interprtations isomorphes sont cart...
Article
Full-text available
Many real-world problems require the ability of reasoning efficiently on formulae which are boolean combinations of boolean and unquantified mathematical propositions. This task requires a fruitful combination of efficient boolean reasoning and mathematical solving capabilities. SAT tools and mathematical reasoners are respectively very effective o...

Network

Cited By