Este trabalho é um estudo inicial sobre a integração de duas áreas da engenharia da computação, as tecnologias adaptativas e os algoritmos genéticos. Para tanto, foi realizada a aplicação de algoritmos genéticos na inferência de autômatos adaptativos. Várias tácnicas foram estudas e propostas para a implementação do algoritmo, visando µa obtenção de resultados cada vez mais satisfatórios. Ambas as tecnologias, algoritmos genéticos e tecnologia adaptativa, possuem caráter fortemente adaptativo, porém com características bastante diferentes na forma que são implementadas e executadas. As inferências, propostas neste trabalho, foram realizadas com sucesso, de maneira que as técnicas descritas podem ser empregadas em ferramentas de auxílio para projetistas desses tipos de dispositivos. Ferramentas que podem vir a ser úteis devido µa complexidade envolvida no desenvolvimento de um autômato adaptativo. Através desta aplicação dos algoritmos genéticos, observando como os autômatos evoluíram durante a execução dos ensaios realizados, acredita-se que foi obtido um entendimento melhor da estrutura e funcionamento dos autômatos adaptativos e de como essas duas tecnologias, tão importantes, podem ser combinadas. / This work is an initial study about the integration of two computing engineering areas, the adaptive technologies and the genetic algorithms. For that, it was per- formed the application of genetic algorithms for the adaptive automata inference. Several techniques were studied and proposed along the algorithm implementation, always seeking for more satisfying results. Both technologies, genetic algorithm and adaptive technology, hold very strong adaptive features, however, with very di®erent characteristics in the way they are implemented and executed. The inferences, proposed in this work, were performed with success, so that the techniques described may be employed in aid tools for designers of such de- vices. Tools that may be useful due to the complexity involved in the development of an adaptive automaton. Through this genetic algorithm application, observing how automata evolved during the algorithm execution, we believe that it was obtained a better under- standing about the adaptive automaton structure and how those two, so impor- tant, technologies can be integrated.

Quand on cherche à situer l'Inférence Grammaticale dans le paysage de la Recherche, on la place volontiers au sein de l'Apprentissage Automatique, qu'on place lui-même volontiers dans le champ de l'Intelligence Artificielle. Ainsi, dans leur livre de référence, Laurent Miclet et Antoine Cornuéjols préfèrent-ils parler d'Apprentissage Artificiel plutôt que d'Apprentissage Automatique, et consacrent-ils un chapitre complet à l'Inférence Grammaticale. C'est l'histoire du Machine Learning qui explique cette hiérarchie. Pourtant, en 2010, elle n'est pas toujours facile à justifier : combien de chercheurs dans le domaine du Machine Learning connaissent-ils le paradigme d'identification à la limite ? Et combien de chercheurs en Inférence Grammaticale maîtrisent-ils la théorie de la régularisation utilisée en optimisation ? Il suffit de suivre des conférences comme ICGI ou ECML pour constater que les communautés sont différentes, tant sur le plan de leurs motivations que sur celui de leurs cultures scientifiques. En outre, lorsqu'on étudie l'histoire des deux domaines, on observe des points de divergence depuis longtemps déjà. D'un autre côté, plusieurs éléments consolident cette hiérarchie. En effet, tous les algorithmes d'identification fournissent in fine des grammaires qui acceptent les données positives et rejettent les données négatives. Donc les grammaires peuvent être vues comme des sortes de classifieurs, et un algorithme d'Inférence Grammaticale comme un apprenant visant à résoudre un problème de classification. De même, le but de l'Inférence Grammaticale Stochastique est d'identifier des distributions de probabilité, et c'est une thématique qu'on retrouve également en Machine Learning. Ainsi, dans ce manuscrit, nous avons choisi d'étudier, à la lumière de nos travaux, les relations entre Inférence Grammaticale et Classification Supervisée.

Proceedings with the revised papers of the FSMNLP (Finite-state Methods and Natural Language Processing) 2007 Workshop in Potsdam / Tagungsband mit den Beiträgen der FSMNLP (Finite-state Methods and Natural Language Processing) 2007 in Potsdam

Les expressions rationnelles et les automates finis sont des objets mathématiques permettant de représenter les langages rationnels. Le lien entre ces structures est le sujet de nombreux thèmes de recherche. Chacun de ces modèles présentent avantages et inconvénients. Nous nous proposons d'établir de nouveaux opérateurs, les multi-tildes-barres, permettant de créer un modèle d'expression se situant entre la structure d'automate et celle d'expression rationnelle simple, utilisant l'union, la concaténation, et l'étoile de Kleene. Les multi-tildes-barres sont basées sur des opérations relativement simples sur les langages, l'ajout et l'élimination du mot vide. Nous étendons les méthodes de conversion classiques entre expressions rationnelles simples et automates finis aux expressions utilisant ces nouveaux opérateurs. Nous montrons également que le pouvoir de factorisation de ces nouvelles expressions est exponentiellement plus grand que celui des expressions rationnelles simples.

In formal language theory, studying shortest strings in languages, and variations thereof, can be useful since these strings can serve as small witnesses for properties of the languages, and can also provide bounds for other problems involving languages. For example, the length of the shortest string accepted by a regular language provides a lower bound on the state complexity of the language. In Chapter 1, we introduce some relevant concepts and notation used in automata and language theory, and we show some basic results concerning the connection between the length of the shortest string and the nondeterministic state complexity of a regular language. Chapter 2 examines the effect of the intersection operation on the length of the shortest string in regular languages. A tight worst-case bound is given for the length of the shortest string in the intersection of two regular languages, and loose bounds are given for two variations on the problem. Chapter 3 discusses languages that are defined over a free group instead of a free monoid. We study the length of the shortest string in a regular language that becomes the empty string in the free group, and a variety of bounds are given for different cases. Chapter 4 mentions open problems and some interesting observations that were made while studying two of the problems: finding good bounds on the length of the shortest squarefree string accepted by a deterministic finite automaton, and finding an efficient way to check if a finite set of finite words generates the free monoid. Some of the results in this thesis have appeared in work that the author has participated in \cite{AngPigRamSha,AngShallit}.

Data-flow analysis is an integral part of any aggressive optimizing compiler. We propose a framework for improving the precision of data-flow analysis in the presence of complex control-flow. We initially perform data-flow analysis to determine those control-flow merges which cause the loss in data-flow analysis precision. The control-flow graph of the program is then restructured such that performing data-flow analysis on the resulting restructured graph gives more precise results. The proposed framework is both simple, involving the familiar notion of product automata, and also general, since it is applicable to any forward or backward data-flow analysis. Apart from proving that our restructuring process is correct, we also show that restructuring is effective in that it necessarily leads to more optimization opportunities. Furthermore, the framework handles the trade-off between the increase in data-flow precision and the code size increase inherent in the restructuring. We show that determining an optimal restructuring is NP-hard, and propose and evaluate a greedy heuristic. The framework has been implemented in the Scale research compiler, and instantiated for the specific problems of Constant Propagation and Liveness analysis. On the SPECINT 2000 benchmark suite we observe an average speedup of 4% in the running times over Wegman-Zadeck conditional constant propagation algorithm and 2% over a purely path profile guided approach for Constant Propagation. For the problem of Liveness analysis, we see an average speedup of 0.8% in the running times over the baseline implementation.

Formal methods offer a way to deal with the complexity of information systems. They are adapted to a variety of domains like design, verification, model-checking, test and supervision. But information systems are also more and more often distributed, first because of the generalization of information networks, but also because inside a single device, like a computer, the numerous components run concurrently. The problem is that concurrency is known to be a major difficulty for the use of formal methods because it causes a combinatorial explosion of the state space of the systems. This difficulty comes sometimes with another one due to time when it plays an important role in the behaviour of the systems, for instance when the execution time is a critical parameter. These two difficulties, concurrency and real-time, have guided my research works. Sometimes I have tackled one of these two aspects separately, but in many of my works, I have dealt with the problems that arise when one studies systems that are both concurrent and real-time. In my habilitation thesis, I give an overview of my recent research works on dependencies between events in real-time distributed systems and on implementability issues for these systems.

In formal language theory, studying shortest strings in languages, and variations thereof, can be useful since these strings can serve as small witnesses for properties of the languages, and can also provide bounds for other problems involving languages. For example, the length of the shortest string accepted by a regular language provides a lower bound on the state complexity of the language. In Chapter 1, we introduce some relevant concepts and notation used in automata and language theory, and we show some basic results concerning the connection between the length of the shortest string and the nondeterministic state complexity of a regular language. Chapter 2 examines the effect of the intersection operation on the length of the shortest string in regular languages. A tight worst-case bound is given for the length of the shortest string in the intersection of two regular languages, and loose bounds are given for two variations on the problem. Chapter 3 discusses languages that are defined over a free group instead of a free monoid. We study the length of the shortest string in a regular language that becomes the empty string in the free group, and a variety of bounds are given for different cases. Chapter 4 mentions open problems and some interesting observations that were made while studying two of the problems: finding good bounds on the length of the shortest squarefree string accepted by a deterministic finite automaton, and finding an efficient way to check if a finite set of finite words generates the free monoid. Some of the results in this thesis have appeared in work that the author has participated in \cite{AngPigRamSha,AngShallit}.

Les travaux présentés ici s'intéressent aux coloriages du plan discret. Ce modèle d'inspiration géométrique est intrinsèquement lié aux modèles de calcul, et son étude se décline ici suivant deux axes complémentaires: calculabilité et combinatoire. Nous montrons en particulier ici comment de nombreux résultats récents s'expriment naturellement à travers le concept de bases, propriétés vérifiées par au moins un point de tout ensemble de coloriages, et d'antibases, contre-exemples à ce concept. Nous examinons ensuite les différents codages du calcul par des jeux de tuiles et exhibons en particulier un nouveau codage épars, permettant de caractériser les degrés Turing des ensembles de coloriages. Enfin nous revenons aux origines en étudiant les pavages du point de vue de la logique. Nous caractérisons ainsi les grandes familles d'ensembles de coloriages par des fragments de la logique monadique du second ordre.

Humans make mistakes, especially when faced to complex tasks, such as the construction of modern hardware or software. This thesis focuses on machine-assisted techniques to guarantee that computers behave correctly. Modern computer systems are large and complex. Automated formal verification stands as an alternative to testing or simulation to ensuring their reliability. It essentially proposes to employ computers to exhaustively check the system behavior. Unfortunately, automated verification suffers from the state-space explosion problem: even relatively small systems can reach a huge number of states. Using the right representation for the system behavior seems to be a key step to tackle the inherent complexity of the problems that automated verification solves. The verification of concurrent systems poses additional issues, as their analysis requires to evaluate, conceptually, all possible execution orders of their concurrent actions. Petri net unfoldings are a well-established verification technique for concurrent systems. They represent behavior by partial orders, which not only is natural but also efficient for automatic verification. This dissertation focuses on the verification of concurrent systems, employing Petri nets to formalize them, and studies two prominent verification techniques: model checking and fault diagnosis. We investigate the unfoldings of Petri nets extended with read arcs. The unfoldings of these so-called contextual nets seem to be a better representation for systems exhibiting concurrent read access to shared resources: they can be exponentially smaller than conventional unfoldings on these cases. Theoretical and practical contributions are made. We first study the construction of contextual unfoldings, introducing algorithms and data structures that enable their efficient computation. We integrate contextual unfoldings with merged processes, another representation of concurrent behavior that alleviates the explosion caused by non-determinism. The resulting structure, called contextual merged processes, is often orders of magnitude smaller than unfoldings, as we experimentally demonstrate. Next, we develop verification techniques based on unfoldings. We define SAT encodings for the reachability problem in contextual unfoldings, thus solving the problem of detecting cycles of asymmetric conflict. Also, an unfolding-based decision procedure for fault diagnosis under fairness constraints is presented, in this case only for conventional unfoldings. Finally, we implement our verification algorithms, aiming at producing a competitive model checker intended to handle realistic benchmarks. We subsequently evaluate our methods over a standard set of benchmarks and compare them with existing unfolding-based techniques. The experiments demonstrate that reachability checking based on contextual unfoldings outperforms existing techniques on a wide number of cases. This suggests that contextual unfoldings, and asymmetric event structures in general, have a rightful place in research on concurrency, also from an efficiency point of view.