SERVICE oriented architecture (SOA) is an architectural style to structure software systems, fostering loose coupling and dynamic integration among the applications. The use of SOA and Web services to develop complex and large business processes demands more formal and systematic testing. In addition, characteristics of this type of software limit the straightforward application of traditional testing techniques. Model-based testing (MBT) is a promising approach to deal with these problems. This dissertation investigates how two modeling techniques, namely Finite State Machine (FSM) and Event Sequence Graph (ESG), can be used to support MBT of service-oriented applications. Both techniques model different aspects and can be applied in a complementary way. Initially, we define an MBT process for service-oriented applications that employs FSMs. Based on previous experience, we propose a model-based approach to test composite services using ESGs. This approach is holistic, once test suites are generated to cover both desired situations (positive testing) and unexpected behaviors (negative testing). Three experimental studies evaluate the proposed approach: (i) a case study, (ii) a cost analysis, and (iii) a study in industry. Testing tools are also presented to support its practical use / A Arquitetura orientada a serviço (SOA) é um estilo arquitetural para estruturar sistemas de software de modo que exista um baixo grau de acoplamento entre as aplicações e essas possam ser facilmente integradas de forma dinâmica. A incorporação de SOA e serviços Web em sistemas que modelam processos de negócios grandes e complexos contribui para a necessidade de testes mais formais e sistemáticos. Além disso, características próprias dessa nova classe de software fazem com que técnicas de teste tradicionais não possam ser diretamente aplicadas. O teste baseado em modelo (TBM) apresenta-se como uma abordagem promissora que busca a resolução desses problemas. Esta tese investiga como duas técnicas de modelagem, Máquina de Estados Finitos (MEF) e Grafo de Sequência de Eventos (GSE), podem ser utilizadas para apoiar o TBM de aplicações orientadas a serviço. Essas técnicas modelam diferentes aspectos e podem ser aplicadas de forma complementar. Inicialmente, é definido um processo de TBM para aplicações orientadas a serviço que emprega MEFs. Com base na experiência adquirida, é proposta uma abordagem baseada em modelo para o teste de serviços compostos usando GSEs. Essa abordagem é holística uma vez que conjuntos de teste são gerados para cobrir tanto situações desejadas (teste positivo) quanto comportamentos inesperados (teste negativo). Três estudos experimentais avaliam a abordagem proposta: (i) um estudo de caso, (ii) uma análise de custo e (ii) um estudo na indústria. Ferramentas de teste também são apresentadas para apoiar o uso prático da abordagem proposta

Dans cette thèse nous proposons un modèle temporel dans le cadre des machines à états abstraits (ASM). Une extension du langage de spécification ASM est développé qui correspond à ce modéle temporel pour le temps continu. L'extension du langage avec des constructions de temps permet de diminuer la taille de la spécification et donc de réduire la probabilité d'erreurs. La sémantique de l'extension du langage ASM est fournie et prend en compte les définitions des fonctions externes, les valeurs des délais et les choix de résolution des non-déterminismes. Un sous-système de vérification des propriétés exprimées en logique FOTL (FirstOrder Timed Logic) est développé. Un simulateur d'ASMs temporisées est développé et implémenté, il comprend un analyseur syntaxique, un interprète du langage, un sous-système de vérification des propriétés ainsi qu'une interface graphique / In this thesis a temporal model for abstract state machines (ASM) method is pro- posed. An extension of ASM specification language on the base of the proposed temporal model with continuous time is developed. The language extension helps to reduce the size of the specification hence to diminish the probability of an error. The semantics of the extended ASM language is developed which takes into account the definitions of external functions, the values of time delays and the method of non-determinism resolving. A subsystem for verification of user properties in the FOTL language is developed. A simulator prototype for ASMs with time is developed and implemented. It includes the parser of the timed ASM language, the interpreter, the verification subsystem and the graphical user interface

In dieser Habilitationsschrift wird die formale Semantik der standardisierten Spezifikationssprache SDL (Specification and Description Language) beschrieben. Da SDL eine sehr umfangreiche Sprache ist, wurde eine repräsentative eingeschränkte Sprache RSDL (Restricted SDL) ausgewählt, um die Konzepte der formalen Definition von SDL darzustellen. Die vorliegende Habilitationsschrift umfaßt zwei große Teile: die Definition der formalen Semantik von RSDL und ihre Implementierung. Die formale Definition der Semantik von RSDL ist verständlich, leicht mit der informalen Beschreibung zu vergleichen und repräsentiert die grundsätzliche Vorstellung von RSDL. Für die Beschreibung werden zwei Teile unterschieden, nämlich die statische Semantik und die dynamische Semantik. Die statische formale Sprachdefinition besteht aus einer konkreten Syntax, einer Menge von Korrektheitsbedingungen, einer Menge von Transformationsregeln und einer abstrakten Syntax als Basis für die dynamische Semantik. Das Ergebnis der statischen Beschreibung ist eine Repräsentation der Spezifikation in abstrakter Syntax. Die Formalisierung der dynamischen Semantik beginnt mit der abstrakten Syntax. Aus dieser abstrakten Syntax wird ein Verhaltensmodell abgeleitet, das auf der mathematischen Theorie der Abstrakten Zustandmaschinen ASM (Abstract State Machines) basiert. Um die Definition der Semantik besonders übersichtlich zu gestalten, wird eine Spezielle Abstrakte Maschine (SAM) unter Nutzung von ASM definiert. Diese abstrakte Maschine stellt eine abstrakte SDL-Maschine dar. Die formale Semantik beschreibt die Eigenschaften von SDL exakt. Um jedoch herauszufinden, ob die Semantik korrekt ist, muß sie mit der Sprachbeschreibung und den Intentionen der Sprachentwickler verglichen werden. Dies geschieht am einfachsten durch eine korrekte Implementierung der Semantik. Die Implementierung der formalen Semantik basiert auf einer Repräsentation der Eingabe als abstrakter Syntaxbaum. Um die Semantik mit minimalem Aufwand zu implementieren, werden existierende Werkzeuge verwendet. Der Compiler wird mit den Standardwerkzeugen lex und yacc generiert. Nach der Syntaxanalyse wird die weitere Verarbeitung über dem abstrakten Syntaxbaum der Eingabe definiert. Die Verarbeitung von abstrakten Syntaxbäumen wird durch ein Werkzeug namens kimwitu erledigt. Mit der hier vorgestellten Technologie wurde die formale Semantik von RSDL implementiert. Entsprechend wird die formale Semantik von SDL implementiert. / In this habilitation thesis the formal semantics of the standardised specification language SDL (Specification and Description Language) is described. Because of the size of the language SDL a representative subset of the language called RSDL (Restricted SDL) was selected to present the concepts of the formal definition. In this thesis two major parts are covered: the definition of the formal semantics and its implementation. The RSDL formal semantics is intelligible, easily comparable with the informal description and represents the general understanding of RSDL. We distinguish between two phases of the definition, namely the static semantics and the dynamic semantics. The static semantics comprises the definition of a concrete grammar, a set of correctness constraints, a set of transformation rules and an abstract syntax as basis for the dynamic semantics. The result of the static semantics is a representation of the specification in abstract syntax. The dynamic semantics starts with the abstract syntax. From here a behaviour model is derived based on the theory of Abstract State Machines (ASM). In order to keep the presentation intelligible a special abstract machine is defined using ASM. This abstract machine in fact represents an abstract SDL-machine. The formal semantics describes the properties of SDL exactly. However, in order to check the correctness of the formalisation, it has to be compared with the informal language description and the intentions of the language designers. This is most easily done using a correct implementation of the semantics. The implementation of the semantics is based on a representation of the input as an abstract syntax tree. For implementing the semantics with minimal effort existing tools are used. The compiler is produced using the standard tools lex and yacc. After parsing the remaining processing is defined over abstract syntax trees, which is covered by a tool called kimwitu. The formal semantics of RSDL is implemented using these tools. The same approach is applicable for SDL.

Made available in DSpace on 2016-12-12T17:38:37Z (GMT). No. of bitstreams: 1 Moacyr Carlos Possan Junior.pdf: 1940795 bytes, checksum: 58824c0ca3ed2180f9e245d34118e117 (MD5) Previous issue date: 2009-12-15 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / This work presents a new methodology for the modeling of supervisory control systems based on state machines with outputs, obtained from the automata which represent the supervisors found with the usage of the Supervisory Control Theory (SCT) of Discrete Event Systems (DES). Procedures like that are useful to create the documentation which represents the control logic for large scale systems and for the code generation, as well as allows that the documentation and program code updates be easier when new requirements to change the control logic in flexible systems appear. Besides, it makes way for the generation of more reliable solutions and also for the possibility of automatic code generation. The proposed technique consists on the obtaining of finite state machines with outputs using as input information the automata of the supervisors obtained by the SCT and the control actions of the system, where the control logic redundancies existing in the model of the automata are eliminated. Methodologies based on either monolithic or local modular approach are proposed, where the obtained machines are further simplified in order to have simpler models which are used as templates for the implementation in Programmable Logic Controller (PLC) using Ladder language. The methodology is shown using a simple manufacturing system as example to help on its understanding. Besides, this work deals with the difficulties found in the migration from the event based theory in the TCS to the signal based practice for the CLPs. After the presentation of this methodology, it is performed the modeling and implementation for a larger system, a manufacturing cell where a comparison with another existing methodology which also has the SCT as base is performed in order to verify the advantages and disadvantages of such methodology. / Este trabalho apresenta uma nova metodologia para a modelagem de sistemas de controle supervisório baseados em máquinas de estados com saídas, obtidas a partir dos autômatos que representam os supervisores encontrados com o uso da Teoria de Controle Supervisório (TCS) de Sistemas a Eventos Discretos (SEDs). Procedimentos como este são úteis para criar a documentação relativa à especificações de sistemas de grande porte e à geração de código, assim como permitem que a atualização da documentação e do código seja facilitada quando surgem novos requisitos para variação da lógica de controle em sistemas flexíveis. Além disso, isso abre espaço para a geração de soluções mais confiáveis e também para a possibilidade de geração automática de código. A técnica proposta consiste em obter máquinas de estados finitos com saídas usando como informação os autômatos dos supervisores obtidos por intermédio da TCS e as ações de controle do sistema, onde redundâncias da lógica de controle presentes no modelo do autômato são eliminadas. São propostas abordagens tanto no contexto monolítico quanto no contexto modular local, onde as máquinas obtidas são reduzidas posteriormente com o intuito de obter modelos mais simples, que servem como referência para a implementação em Controlador Lógico Programável (CLP) usando linguagem Ladder. A metodologia é demonstrada usando um sistema de manufatura simples como exemplo a fim de facilitar sua compreensão. Além disso, este trabalho trata das dificuldades encontradas na migração da teoria baseada em eventos da TCS na prática baseada em sinais dos CLPs. Após a apresentação da metodologia, é feita a modelagem e implementação para um sistema de maior porte, uma célula de manufatura onde é efetuada uma comparação com uma metodologia já existente que também têm como base a TCS, com o intuito de verificar as vantagens e desvantagens dessa metodologia

SERVICE oriented architecture (SOA) is an architectural style to structure software systems, fostering loose coupling and dynamic integration among the applications. The use of SOA and Web services to develop complex and large business processes demands more formal and systematic testing. In addition, characteristics of this type of software limit the straightforward application of traditional testing techniques. Model-based testing (MBT) is a promising approach to deal with these problems. This dissertation investigates how two modeling techniques, namely Finite State Machine (FSM) and Event Sequence Graph (ESG), can be used to support MBT of service-oriented applications. Both techniques model different aspects and can be applied in a complementary way. Initially, we define an MBT process for service-oriented applications that employs FSMs. Based on previous experience, we propose a model-based approach to test composite services using ESGs. This approach is holistic, once test suites are generated to cover both desired situations (positive testing) and unexpected behaviors (negative testing). Three experimental studies evaluate the proposed approach: (i) a case study, (ii) a cost analysis, and (iii) a study in industry. Testing tools are also presented to support its practical use / A Arquitetura orientada a serviço (SOA) é um estilo arquitetural para estruturar sistemas de software de modo que exista um baixo grau de acoplamento entre as aplicações e essas possam ser facilmente integradas de forma dinâmica. A incorporação de SOA e serviços Web em sistemas que modelam processos de negócios grandes e complexos contribui para a necessidade de testes mais formais e sistemáticos. Além disso, características próprias dessa nova classe de software fazem com que técnicas de teste tradicionais não possam ser diretamente aplicadas. O teste baseado em modelo (TBM) apresenta-se como uma abordagem promissora que busca a resolução desses problemas. Esta tese investiga como duas técnicas de modelagem, Máquina de Estados Finitos (MEF) e Grafo de Sequência de Eventos (GSE), podem ser utilizadas para apoiar o TBM de aplicações orientadas a serviço. Essas técnicas modelam diferentes aspectos e podem ser aplicadas de forma complementar. Inicialmente, é definido um processo de TBM para aplicações orientadas a serviço que emprega MEFs. Com base na experiência adquirida, é proposta uma abordagem baseada em modelo para o teste de serviços compostos usando GSEs. Essa abordagem é holística uma vez que conjuntos de teste são gerados para cobrir tanto situações desejadas (teste positivo) quanto comportamentos inesperados (teste negativo). Três estudos experimentais avaliam a abordagem proposta: (i) um estudo de caso, (ii) uma análise de custo e (ii) um estudo na indústria. Ferramentas de teste também são apresentadas para apoiar o uso prático da abordagem proposta

Dans une composition de services Web basée protocole, un ensemble de services composants se collaborent pour donner lieu à un service Composite. Chaque service est représenté par un automate à états finis (AEF). Au sein d’un AEF, chaque transition exprime l’exécution d’une opération qui fait avancer le service vers un état suivant. Une exécution du composite correspond à une séquence de transitions où chacune est déléguée à un des composants. Lors de l’exécution du composite, un ou plusieurs composants peuvent devenir indisponibles. Ceci peut produire une exécution incomplète du composite, et de ce fait un recouvrement est nécessaire. Le recouvrement consiste à transformer l’exécution incomplète en une exécution alternative ayant encore la capacité d’aller vers un état final. La transformation s'effectue en compensant certaines transitions et exécutant d’autres. Cette thèse présente une étude formelle du problème de recouvrement dans une composition de service Web basée protocole. Le problème de recouvrement consiste à trouver une meilleure exécution alternative parmi celles disponibles. Une meilleure alternative doit être atteignable à partir de l’exécution incomplète avec un nombre minimal de compensations visibles (vis-à-vis le client). Pour une exécution alternative donnée, nous prouvons que le problème de décision associé au calcul du nombre de transitions invisiblement compensées est NP-Complet. De ce fait, nous concluons que le problème de décision associé au recouvrement appartient à la classe ΣP2. / In a protocol-based Web service composition, a set of available component services collaborate together in order to provide a new composite service. Services export their protocols as finite state machines (FSMs). A transition in the FSM represents a task execution that makes the service moving to a next state. An execution of the composite corresponds to a sequence of transitions where each task is delegated to a component service. During composite run, one or more delegated components may become unavailable due to hard or soft problems on the Network. This unavailability may result in a failed execution of the composite. We provide in this thesis a formal study of the automatic recovery problem in the protocol-based Web service composition. Recovery consists in transforming the failed execution into a recovery execution. Such a transformation is performed by compensating some transitions and executing some others. The recovery execution is an alternative execution of the composite that still has the ability to reach a final state. The recovery problem consists then in finding the best recovery execution(s) among those available. The best recovery execution is attainable from the failed execution with a minimal number of visible compensations with respect to the client. For a given recovery execution, we prove that the decision problem associated with computing the number of invisibly-compensated transitions is NP-complete. Thus, we conclude that deciding of the best recovery execution is in ΣP2.

This thesis considers control systems governed by autonomous decision-makers and humans. We formalise and compute low-complex control policies with applications to large-scale systems, and propose human interaction models for controllers to compute interaction-aware decisions. In the first part of the thesis, we consider complexity-aware decision-making, formalising the complexity of control policies and constructing algorithms that compute low-complexity control policies. More precisely, first, we consider large-scale control systems given by hierarchical finite state machines (HFSMs) and present a planning algorithm for such systems that exploits the hierarchy to compute optimal policies efficiently. The algorithm can also handle changes in the system with ease. We prove these properties and conduct simulations on HFSMs with up to 2 million states, including a robot application, where our algorithm outperforms both Dijkstra's algorithm and Contraction Hierarchies.  Second, we present a planning objective for control systems modelled as finite state machines yielding an explicit trade-off between a policy's performance and complexity. We consider Kolmogorov complexity since it captures the ultimate compression of an object on a universal Turing machine. We prove that this trade-off is hard to optimise in the sense that dynamic programming is infeasible. Nonetheless, we present two heuristic algorithms obtaining low-complexity policies and evaluate the algorithms on a simple navigation task for a mobile robot, where we obtain low-complexity policies that concur with intuition.  In the second part of the thesis, we consider human-in-the-loop systems and predict human decision-making in such systems. First, we look at how the interaction between a robot and a human in a control system can be predicted using game theory, focusing on an autonomous truck platoon interacting with a human-driven car. The interaction is modelled as a hierarchical dynamic game, where the hierarchical decomposition is temporal with a high-fidelity tactical horizon predicting immediate interactions and a low-fidelity strategic horizon estimating long-term behaviour. The game enables feasible computations validated through simulations yielding situation-aware behaviour with natural and safe interactions.  Second, we seek models to explain human decision-making, focusing on driver overtaking scenarios. The overtaking problem is formalised as a decision problem with perceptual uncertainty. We propose and numerically analyse risk-agnostic and risk-aware decision models, judging if an overtaking is desirable. We show how a driver's decision time and confidence level can be characterised through two model parameters, which collectively represent human risk-taking behaviour. We detail an experimental testbed for evaluating the decision-making process in the overtaking scenario and present some preliminary experimental results from two human drivers. / Denna avhandling studerar styrsystem med autonoma beslutsfattare och människor. Vi formaliserar och beräknar styrlagar av låg komplexitet med tillämpningar på storskaliga system samt föreslår modeller för mänsklig interaktion som kan användas av regulatorer för att beräkna interaktionsmedvetna beslut. I den första delen av denna avhandling studerar vi komplexitet-medveten beslutsfattning, där vi formaliserar styrlagars komplexitet samt konstruerar algoritmer som beräknar styrlagar med låg komplexitet. Mer precist, först studerar vi storskaliga system givna av hierarkiska finita tillståndsmaskiner (HFSMs) och presenterar en planeringsalgoritm för sådana system som utnyttjar hierarkin för att beräkna optimala styrlagar effektivt. Algoritmen kan också lätt hantera förändringar i systemet. Vi bevisar dessa egenskaper och utför simuleringar på HFSMs med upp till 2 miljoner tillstånd, inklusive en robot-applikation, där vår algorithm överträffar både Dijkstra's algoritm och så kallade Contraction Hierarchies. För det andra så presenterar vi ett planeringsobjektiv för finita tillståndsmaskiner som ger en explicit avvägning mellan ett styrlags prestanda och komplexitet. Vi använder Kolmogorovkomplexitet då den fångar den ultimata komprimeringen av ett objekt i en universell Turing-maskin. Vi bevisar att detta objektiv är icke-trivial att optimera över i avseendet att dynamisk programming är omöjligt att utföra. Vi presenterar två algoritmer som beräknar styrlagar med låg komplexitet och evaluerar våra algoritmer på ett enkelt navigationsproblem där vi erhåller styrlagar av låg komplexitet som instämmer med intuition. I den andra delen av denna avhandling behandlar vi reglersystem där en människa interagerar med systemet och studerar hur mänskligt beslutsfattande i sådana system kan förutspås. Först studerar vi hur interaktionen mellan en maskin och en människa i ett reglersystem can förutspås med hjälp av spelteori, med fokus på en självkörande lastbilskonvoj som interagerar med en mänskligt styrd bil. Interaktionen är modellerad som ett hierarkiskt dynamiskt spel, där den hierarkiska indelningen är tidsmässig med en högupplöst taktil horisont som förutspår omedelbara interaktioner samt en lågupplöst strategisk horisont som estimerar långtgående interaktioner. Indelning möjliggör beräkningar som vi validerar via simuleringar där vi får situations-medvetet beteende med naturliga och säkra interaktioner. För det andra söker vi en model med få parametrar som förklarar mänskligt beteende där vi fokuserar på omkörningar. Vi formaliserar omkörningsproblemet som ett beslutfattningsproblem med perceptuell osäkerhet. Vi presenterar och analyserar numeriskt risk-agnostiska och risk-medvetna beslutsmodeller som avväger om en omkörning är önskvärd. Vi visar hur en förares beslutstid och konfidensnivå kan karakteriserar via två modellparametrar som tillsammans representerar mänskligt risk-beteende. Vi beskriver en experimentell testbädd och presentar preliminära resultat från två mänskliga förare. / <p>QC 20230523</p>

Les fichiers sons qui accompagne mon document sont au format midi. Le programme que nous avons développés pour ce travail est en language Python. / Les réseaux de neurones à réservoir, dont le principe est de combiner un vaste réseau de neurones fixes avec un apprenant ne possédant aucune forme de mémoire, ont récemment connu un gain en popularité dans les communautés d’apprentissage machine, de traitement du signal et des neurosciences computationelles. Ces réseaux qui peuvent être classés en deux catégories : 1. les réseaux à états échoïques (ESN)[29] dont les activations des neurones sont des réels 2. les machines à états liquides (LSM)[43] dont les neurones possèdent des potentiels d’actions, ont été appliqués à différentes tâches [11][64][49][45][38] dont la génération de séquences mélodiques [30]. Dans le cadre de la présente recherche, nous proposons deux nouveaux modèles à base de réseaux de neurones à réservoir. Le premier est un modèle pour la reconnaissance de rythmes utilisant deux niveaux d’apprentissage, et avec lequel nous avons été en mesure d’obtenir des résultats satisfaisants tant au niveau de la reconnaissance que de la résistance au bruit. Le second modèle sert à l’apprentissage et à la génération de séquences périodiques. Ce modèle diffère du modèle génératif classique utilisé avec les ESN à la fois au niveau de ses entrées, puisqu’il possède une Horloge, ainsi qu’au niveau de l’algorithme d’apprentissage, puisqu’il utilise un algorithme que nous avons spécialement développé pour cette tache et qui se nomme "Orbite". La combinaison de ces deux éléments, nous a permis d’obtenir de bons résultats, pour la génération, le sur-apprentissage et l’extraction de données. Nous pensons également que ce modèle ouvre une fenêtre intéressante vers la réalisation d’un orchestre entièrement virtuel et nous proposons deux architectures possibles que pourrait avoir cet orchestre. Dans la dernière partie de ce travail nous présentons les outils que nous avons développés pour faciliter notre travail de recherche. / Reservoir computing, the combination of a recurrent neural network and one or more memoryless readout units, has seen recent growth in popularity in and machine learning, signal processing and computational neurosciences. Reservoir-based methods have been successfully applied to a wide range of time series problems [11][64][49][45][38] including music [30], and usually can be found in two flavours: Echo States Networks(ESN)[29], where the reservoir is composed of mean rates neurons, and Liquid Sates Machines (LSM),[43] where the reservoir is composed of spiking neurons. In this work, we propose two new models based upon the ESN architecture. The first one is a model for rhythm recognition that uses two levels of learning and with which we have been able to get satisfying results on both recognition and noise resistance. The second one is a model for learning and generating periodic sequences, with this model we introduced a new architecture for generative models based upon ESNs where the reservoir receives inputs from a clock, as well as a new learning algorithm that we called "Orbite". By combining these two elements within our model, we were able to get good results on generation, over-fitting and data extraction. We also believe that a combination of several instances of our model can serve as a basis for the elaboration of an entirely virtual orchestra, and we propose two architectures that this orchestra may have. In the last part of this work, we briefly present the tools that we have developed during our research.

<p>Deep ‘Analog Artificial Neural Networks’ (AANNs) perform complex classification problems with high accuracy. However, they rely on humongous amount of power to perform the calculations, veiling the accuracy benefits. The biological brain on the other hand is significantly more powerful than such networks and consumes orders of magnitude less power, indicating some conceptual mismatch. Given that the biological neurons are locally connected, communicate using energy efficient trains of spikes, and the behavior is non-deterministic, incorporating these effects in Artificial Neural Networks (ANNs) may drive us few steps towards a more realistic neural networks. </p> <p> </p> <p>Emerging devices can offer a plethora of benefits including power efficiency, faster operation, low area in a vast array of applications. For example, memristors and Magnetic Tunnel Junctions (MTJs) are suitable for high density, non-volatile Random Access Memories when compared with CMOS implementations. In this work, we analyze the possibility of harnessing the characteristics of such emerging devices, to achieve neuro-inspired solutions to intricate problems.</p> <p> </p> <p>We propose how the inherent stochasticity of nano-scale resistive devices can be utilized to realize the functionality of spiking neurons and synapses that can be incorporated in deep stochastic Spiking Neural Networks (SNN) for image classification problems. While ANNs mainly dwell in the aforementioned classification problem solving domain, they can be adapted for a variety of other applications. One such neuro-inspired solution is the Cellular Neural Network (CNN) based Boolean satisfiability solver. Boolean satisfiability (k-SAT) is an NP-complete (k≥3) problem that constitute one of the hardest classes of constraint satisfaction problems. We provide a proof of concept hardware based analog k-SAT solver that is built using MTJs. The inherent physics of MTJs, enhanced by device level modifications, is harnessed here to emulate the intricate dynamics of an analog, CNN based, satisfiability (SAT) solver. </p> <p> </p> <p>Furthermore, in the effort of reaching human level performance in terms of accuracy, increasing the complexity and size of ANNs is crucial. Efficient algorithms for evaluating neural network performance is of significant importance to improve the scalability of networks, in addition to designing hardware accelerators. We propose a scalable approach for evaluating Liquid State Machines: a bio-inspired computing model where the inputs are sparsely connected to a randomly interlinked reservoir (or liquid). It has been shown that biological neurons are more likely to be connected to other neurons in the close proximity, and tend to be disconnected as the neurons are spatially far apart. Inspired by this, we propose a group of locally connected neuron reservoirs, or an ensemble of liquids approach, for LSMs. We analyze how the segmentation of a single large liquid to create an ensemble of multiple smaller liquids affects the latency and accuracy of an LSM. In our analysis, we quantify the ability of the proposed ensemble approach to provide an improved representation of the input using the Separation Property (SP) and Approximation Property (AP). Our results illustrate that the ensemble approach enhances class discrimination (quantified as the ratio between the SP and AP), leading to improved accuracy in speech and image recognition tasks, when compared to a single large liquid. Furthermore, we obtain performance benefits in terms of improved inference time and reduced memory requirements, due to lower number of connections and the freedom to parallelize the liquid evaluation process.</p>

Les réseaux de neurones à réservoir, dont le principe est de combiner un vaste réseau de neurones fixes avec un apprenant ne possédant aucune forme de mémoire, ont récemment connu un gain en popularité dans les communautés d’apprentissage machine, de traitement du signal et des neurosciences computationelles. Ces réseaux qui peuvent être classés en deux catégories : 1. les réseaux à états échoïques (ESN)[29] dont les activations des neurones sont des réels 2. les machines à états liquides (LSM)[43] dont les neurones possèdent des potentiels d’actions, ont été appliqués à différentes tâches [11][64][49][45][38] dont la génération de séquences mélodiques [30]. Dans le cadre de la présente recherche, nous proposons deux nouveaux modèles à base de réseaux de neurones à réservoir. Le premier est un modèle pour la reconnaissance de rythmes utilisant deux niveaux d’apprentissage, et avec lequel nous avons été en mesure d’obtenir des résultats satisfaisants tant au niveau de la reconnaissance que de la résistance au bruit. Le second modèle sert à l’apprentissage et à la génération de séquences périodiques. Ce modèle diffère du modèle génératif classique utilisé avec les ESN à la fois au niveau de ses entrées, puisqu’il possède une Horloge, ainsi qu’au niveau de l’algorithme d’apprentissage, puisqu’il utilise un algorithme que nous avons spécialement développé pour cette tache et qui se nomme "Orbite". La combinaison de ces deux éléments, nous a permis d’obtenir de bons résultats, pour la génération, le sur-apprentissage et l’extraction de données. Nous pensons également que ce modèle ouvre une fenêtre intéressante vers la réalisation d’un orchestre entièrement virtuel et nous proposons deux architectures possibles que pourrait avoir cet orchestre. Dans la dernière partie de ce travail nous présentons les outils que nous avons développés pour faciliter notre travail de recherche. / Reservoir computing, the combination of a recurrent neural network and one or more memoryless readout units, has seen recent growth in popularity in and machine learning, signal processing and computational neurosciences. Reservoir-based methods have been successfully applied to a wide range of time series problems [11][64][49][45][38] including music [30], and usually can be found in two flavours: Echo States Networks(ESN)[29], where the reservoir is composed of mean rates neurons, and Liquid Sates Machines (LSM),[43] where the reservoir is composed of spiking neurons. In this work, we propose two new models based upon the ESN architecture. The first one is a model for rhythm recognition that uses two levels of learning and with which we have been able to get satisfying results on both recognition and noise resistance. The second one is a model for learning and generating periodic sequences, with this model we introduced a new architecture for generative models based upon ESNs where the reservoir receives inputs from a clock, as well as a new learning algorithm that we called "Orbite". By combining these two elements within our model, we were able to get good results on generation, over-fitting and data extraction. We also believe that a combination of several instances of our model can serve as a basis for the elaboration of an entirely virtual orchestra, and we propose two architectures that this orchestra may have. In the last part of this work, we briefly present the tools that we have developed during our research. / Les fichiers sons qui accompagne mon document sont au format midi. Le programme que nous avons développés pour ce travail est en language Python.