Designing and developing complex engineering systems is a collaborative effort. In Model-Based Systems Engineering (MBSE), this collaboration is supported through the use of formal, computer-interpretable models, allowing stakeholders to address concerns using well-defined modeling languages. However, because concerns cannot be separated completely, implicit relationships and dependencies among the various models describing a system are unavoidable. Given that models are typically co-evolved and only weakly integrated, inconsistencies in the agglomeration of the information and knowledge encoded in the various models are frequently observed. The challenge is to identify such inconsistencies in an automated fashion. In this research, a probabilistic (Bayesian) approach to abductive reasoning about the existence of specific types of inconsistencies and, in the process, semantic overlaps (relationships and dependencies) in sets of heterogeneous models is presented. A prior belief about the manifestation of a particular type of inconsistency is updated with evidence, which is collected by extracting specific features from the models by means of pattern matching. Inference results are then utilized to improve future predictions by means of automated learning. The effectiveness and efficiency of the approach is evaluated through a theoretical complexity analysis of the underlying algorithms, and through application to a case study. Insights gained from the experiments conducted, as well as the results from a comparison to the state-of-the-art have demonstrated that the proposed method is a significant improvement over the status quo of inconsistency identification in MBSE.

Feature extraction is an extremely important pre-processing step to pattern recognition, and machine learning problems. This thesis highlights how one can best extract features from the data in an exhaustively online and purely adaptive manner. The solution to this problem is given for both labeled and unlabeled datasets, by presenting a number of novel on-line learning approaches. Specifically, the differential equation method for solving the generalized eigenvalue problem is used to derive a number of novel machine learning and feature extraction algorithms. The incremental eigen-solution method is used to derive a novel incremental extension of linear discriminant analysis (LDA). Further the proposed incremental version is combined with extreme learning machine (ELM) in which the ELM is used as a preprocessor before learning. In this first key contribution, the dynamic random expansion characteristic of ELM is combined with the proposed incremental LDA technique, and shown to offer a significant improvement in maximizing the discrimination between points in two different classes, while minimizing the distance within each class, in comparison with other standard state-of-the-art incremental and batch techniques. In the second contribution, the differential equation method for solving the generalized eigenvalue problem is used to derive a novel state-of-the-art purely incremental version of slow feature analysis (SLA) algorithm, termed the generalized eigenvalue based slow feature analysis (GENEIGSFA) technique. Further the time series expansion of echo state network (ESN) and radial basis functions (EBF) are used as a pre-processor before learning. In addition, the higher order derivatives are used as a smoothing constraint in the output signal. Finally, an online extension of the generalized eigenvalue problem, derived from James Stone’s criterion, is tested, evaluated and compared with the standard batch version of the slow feature analysis technique, to demonstrate its comparative effectiveness. In the third contribution, light-weight extensions of the statistical technique known as canonical correlation analysis (CCA) for both twinned and multiple data streams, are derived by using the same existing method of solving the generalized eigenvalue problem. Further the proposed method is enhanced by maximizing the covariance between data streams while simultaneously maximizing the rate of change of variances within each data stream. A recurrent set of connections used by ESN are used as a pre-processor between the inputs and the canonical projections in order to capture shared temporal information in two or more data streams. A solution to the problem of identifying a low dimensional manifold on a high dimensional dataspace is then presented in an incremental and adaptive manner. Finally, an online locally optimized extension of Laplacian Eigenmaps is derived termed the generalized incremental laplacian eigenmaps technique (GENILE). Apart from exploiting the benefit of the incremental nature of the proposed manifold based dimensionality reduction technique, most of the time the projections produced by this method are shown to produce a better classification accuracy in comparison with standard batch versions of these techniques - on both artificial and real datasets.

No description available.

Lorsqu’on fait face à des systèmes non linéaires, les notions classiques de stabilité ne suffisent pas à garantir un comportement approprié vis-à-vis de problématiques telles que le suivi de trajectoires, la synchronisation et la conception d’observateurs. La stabilité incrémentale a été proposée en tant qu’outil permettant de traiter de tels problèmes et de garantir que le système présente des comportements qualitatifs pertinents. Cependant, comme c’est souvent le cas avec les systèmes non linéaires, la complexité de l’analyse conduit les ingénieurs à rechercher des relaxations, ce qui introduit du conservatisme. Dans cette thèse, nous nous intéressons à la stabilité incrémentale d’une classe spécifique de systèmes, à savoir les systèmes affines par morceaux, qui pourraient fournir un outil avantageux pour aborder la stabilité incrémentale de systèmes dynamiques plus génériques.Les systèmes affines par morceaux ont un espace d’états partitionné, et sa dynamique dans chaque région est régie par une équation différentielle affine. Ils peuvent représenter des systèmes contenant des non linéarités affines par morceaux, ainsi que servir comme des approximations de systèmes non linéaires plus génériques. Ce qui est plus important, leur description est relativement proche de celle des systèmes linéaires, ce qui permet d’obtenir des conditions d’analyse exprimées comme des inégalités matricielles linéaires qui peuvent être traités numériquement de façon efficace par des solveurs existants.Dans la première partie de ce document de thèse, nous passons en revue la littérature sur l’analyse des systèmes affines par morceaux en utilisant des techniques de Lyapunov et la dissipativité. Nous proposons ensuite de nouvelles conditions pour l’analyse du gain L2 incrémental et la stabilité asymptotique incrémentale des systèmes affines par morceaux exprimés en tant qu’inégalités matricielles linéaires. Ces conditions sont montrées être moins conservatives que les résultats précédents et sont illustrées par des exemples numériques.Dans la deuxième partie, nous considérons le cas des systèmes affines par morceaux incertains représentés comme l’interconnexion entre un système nominal et un bloc d’incertitude structuré. En utilisant la théorie de la séparation des graphes, nous proposons des conditions qui étendent le cadre des contraintes quadratiques intégrales afin de considérer le cas où le système nominal est affine par morceaux, à la fois dans les cas non incrémental et incrémental. Via la théorie de la dissipativité, ces conditions sont ensuite exprimées en tant qu’inégalités matricielles linéaires.Finalement, la troisième partie de ce document de thèse est consacrée à l’analyse de systèmes non linéaires de Lur’e incertains. Nous développons une nouvelle technique d’approximation permettant de réécrire ces systèmes de façon équivalente comme des systèmes affines par morceaux incertains connectés avec l’erreur d’approximation. L’approche proposée garantit que l’erreur d’approximation est Lipschitz continue avec la garantie d’une borne supérieure prédéterminée sur la constante de Lipschitz. Cela nous permet d’utiliser les techniques susmentionnées pour analyser des classes plus génériques de systèmes non linéaires. / When dealing with nonlinear systems, regular notions of stability are not enough to ensure an appropriate behavior when dealing with problems such as tracking, synchronization and observer design. Incremental stability has been proposed as a tool to deal with such problems and ensure that the system presents relevant qualitative behavior. However, as it is often the case with nonlinear systems, the complexity of the analysis leads engineers to search for relaxations, which introduce conservatism. In this thesis, we focus on the incremental stability of a specific class of systems, namely piecewise-affine systems, which could provide a valuable tool for approaching the incremental stability of more general dynamical systems.Piecewise-affine systems have a partitioned state space, in each region of which the dynamics are governed by an affine differential equation. They can represent systems containing piecewise-affine nonlinearities, as well as serve as approximations of more general nonlinear systems. More importantly, their description is relatively close to that of linear systems, allowing us to obtain analysis conditions expressed as linear matrix inequalities that can be efficiently handled numerically by existing solvers.In the first part of this memoir, we review the literature on the analysis of piecewise-affine systems using Lyapunov and dissipativity techniques. We then propose new conditions for the analysis of incremental L2-gain and incremental asymptotic stability of piecewise-affine systems expressed as linear matrix inequalities. These conditions are shown to be less conservative than previous results and illustrated through numerical examples.In the second part, we consider the case of uncertain piecewise-affine systems represented as the interconnection between a nominal system and a structured uncertainty block. Using graph separation theory, we propose conditions that extend the framework of integral quadratic constraints to consider the case when the nominal system is piecewise affine, both in the non-incremental and incremental cases. Through dissipativity theory, these conditions are then expressed as linear matrix inequalities.Finally, the third part of this memoir is devoted to the analysis of uncertain Lur’e-type nonlinear systems. We develop a new approximation technique allowing to equivalently rewrite such systems as uncertain piecewise-affine systems connected with the approximation error. The proposed approach ensures that the approximation error is Lipschitz continuous with a guaranteed pre-specified upper bound on the Lipschitz constant. This enables us to use the aforementioned techniques to analyze more general classes of nonlinear systems.

O propósito do presente estudo foi obter evidências de validade e reprodutibilidade dos limiares de variabilidade da frequência cardíaca (VFC) durante a corrida. Dezenove sujeitos homens, saudáveis e praticantes de corrida (30,4 ± 4,1 anos; 175,9 ± 6,4 cm; 74,3 ± 8,5 kg) foram submetidos a um teste progressivo máximo em esteira rolante com velocidade inicial 5 km.h-1 e incrementos de 1 km.h-1 a cada 3 minutos (1% de inclinação constante) até exaustão voluntária. Todos os indivíduos realizaram o reteste em um intervalo de tempo entre 48 horas e uma semana. Foram realizadas as medidas das trocas gasosas, do lactato sanguíneo e da VFC (plotagem de Poincaré). Os limiares aeróbio (LAe) e anaeróbio (LAn) foram determinados pelos limares de lactato, ventilatórios e da VFC. Para a comparação entre os métodos foi uitlizada ANOVA para medidas repetidas, acompanhada de teste de post hoc de Bonferroni. A reprodutibilidade das variáveis analisadas foram verificadas pela plotagem de Bland-Altman e pelo coeficiente de correlação intraclasse (CCI). Os resultados do presente estudo demonstraram que a velocidade correspondente ao segundo e terceiro modelos utilizados para se determinar o LA pela VFC não eram significativamente diferentes (p > 0,05) do primeiro limiar de lactato e ventilatório. Em relação ao LAn, não foram observadas diferenças significativas nas velocidades correspondentes ao LAn detectado pelos diferentes métodos (p > 0,05). Os valores do CCI estavam entre 0,69 a 0,80 (p < 0,001). Conclui-se que o LAe e o LAn podem ser identificados pela análise da VFC, desde que se utilize os procedimentos empregados na presente investigação / The aim of the present study was to obtain evidences of validity and reliability of the thresholds of heart rate variability (HRV). Nineteen male subjects, healthy and runners (30,4 ± 4,1 years; 175,9 ± 6,4 cm; 74,3 ± 8,5 kg) performed a progressive maximal test on a treadmill with initial velocity 5 km.h-1 e increases of 1 km.h-1 every 3 minutes (1% slope) until voluntary exhaustion. All subjects performed the retest at an interval of time between 48 hours and one week. It was measured gas exchange, blood lactate and heart rate variability (Poincaré plot). The aerobic threshold (AT) and anaerobic (AnT) were determined by lactate, ventilatory and heart rate variability. ANOVA for repeated measures and post-hoc test of Bonferroni was used to compare the methods. To analyze the reproducibility of the variables were used the Bland- Altman plots and intraclass correlation coefficient (ICC). The results of this study show that the velocity at the second and third models employed to determine the AT by HRV were not significantly different (p > 0.05) of the first lactate threshold and ventilatory. Similarly, there were no significant differences in the velocities corresponding to AnT detected by different methods (p> 0.05). The ICC values were between 0.69 to 0.80 (p < 0.001). We conclude that the AT and the AnT can be estimated by HRV analysis, since it utilizes the procedures employed in this study

Die Studie hat zum Ziel, im Rahmen des Äußerungsproduktionsmodells die Annahme thinking for speaking (Slobin, 1996) experimentell zu überprüfen. Ansatzpunkt ist der Unterschied in der Stellung der Markierung einer Entscheidungsfrage zwischen drei Sprachen. Während der Fragemodus im Deutschen/Polnischen vor dem Frageinhalt markiert wird, geschieht die Modusmarkierung im Chinesischen nach dem Inhalt. Um die entsprechende Satzstruktur aufzubauen, sollte der Fragemodus beim syntaktischen Kodierungsprozess (speaking) im Deutschen/Polnischen vor, im Chinesischen nach dem Inhalt verarbeitet werden. Unter Zugrundelegung der inkrementellen Äußerungsproduktion gehen wir davon aus, dass die Verarbeitungsreihenfolge beim vorangehenden Konzeptualisierungsprozess (thinking) mit der syntaktischen Kodierungsabfolge übereinstimmt: Der Fragemodus wird im Deutschen/ Polnischen vor, im Chinesischen hingegen nach dem Inhalt konzeptualisiert. Um den zeitlichen Ablauf der zwei Konzeptualisierungsprozess zu ermitteln, bedienten wir uns des lateralisierten Bereitschaftspotenzials (lateralized readiness potential, LRP) im binären Wahlreaktions-Go/Nogo-Paradigma. Im Versuch reagierten deutsche, chinesische und polnische Muttersprachler auf dargebotene Bilder mit Tastendrücken und Sprechen. Zu beobachten war das Auftreten eines LRP bei Nogo, das signalisierte, in welcher Reihenfolge sich die Handwahl und die Nogo-Entscheidung realisierten Ein Nogo-LRP trat bei allen drei Sprechergruppen auf. Zudem wurde festgestellt, dass die Nogo-Entscheidung, die erwartungsgemäß mit der sprachlichen Verarbeitung des Fragemodus verbunden wurde, bei den drei Sprechergruppen ungefähr zeitgleich getroffen wurde. Die Befunde legen nahe, dass der Fragemodus in den drei Sprachen zeitlich nicht unterschiedlich, sondern einheitlich geplant wurde. Die Schlussfolgerung wird im Rahmen von thinking for speaking sowie dem Äußerungsproduktionsmodell diskutiert. / Languages differ in the marking of the sentence mood of a polar interrogative (yes/no question). For instance, the interrogative mood is marked at the beginning of the surface structure in Polish, whereas the marker appears at the end in Chinese. In order to generate the corresponding sentence frame, the syntactic specification of the interrogative mood is early in Polish and late in Chinese. In this respect, German belongs to an interesting intermediate class. The yes/no-question is expressed by a shift of the finite verb from its final position in the underlying structure into the utterance initial position, a move affecting, hence, both the sentence final and the sentence initial constituents. The present study aimed to investigate whether during generation of the semantic structure of a polar interrogative, i.e. the processing preceding the grammatical formulation, the interrogative mood is encoded according to its position in the syntactic structure at distinctive time points in Chinese, German, and Polish. In a two-choice go/nogo experimental design, native speakers of the three languages responded to pictures by pressing buttons and producing utterances in their native language while their brain potentials were recorded. The emergence and latency of lateralized readiness potentials (LRP) in nogo conditions, in which speakers asked a yes/no question, should indicate the time point of processing the interrogative mood. The results revealed that Chinese, German, and Polish native speakers did not differ from each other in the electrophysiological indicator. The findings suggest that the semantic encoding of the interrogative mood is temporally consistent across languages despite its disparate syntactic specification. The consistent encoding may be ascribed to economic processing of interrogative moods at various sentential positions of the syntactic structures in languages or, more generally, to the overarching status of sentence mood in the semantic structure.

No description available.

Nous présentons deux contributions majeures : 1) une combinaison de plusieurs descripteurs d’images pour la classification à grande échelle, 2) des algorithmes parallèles de SVM pour la classification d’images à grande échelle. Nous proposons aussi un algorithme incrémental et parallèle de classification lorsque les données ne peuvent plus tenir en mémoire vive. / We have proposed a novel method of combination multiple of different features for image classification. For large scale learning classifiers, we have developed the parallel versions of both state-of-the-art linear and nonlinear SVMs. We have also proposed a novel algorithm to extend stochastic gradient descent SVM for large scale learning. A class of large scale incremental SVM classifiers has been developed in order to perform classification tasks on large datasets with very large number of classes and training data can not fit into memory.

A segmentação de imagens é um problema muito importante em visão computacional, no qual uma imagem é dividida em regiões relevantes, tal como para isolar objetos de interesse de uma dada aplicação. Métodos de segmentação baseados na transformada imagem-floresta (IFT, Image Foresting Transform), com funções de conexidade monotonicamente incrementais (MI) têm alcançado um grande sucesso em vários contextos. Na segmentação interativa de imagens, na qual o usuário pode especificar o objeto desejado, novas sementes podem ser adicionadas e/ou removidas para corrigir a rotulação até conseguir a segmentação esperada. Este processo gera uma sequência de IFTs que podem ser calculadas de modo mais eficiente pela DIFT (Differential Image Foresting Transform). Recentemente, funções de conexidade não monotonicamente incrementais (NMI) têm sido usadas com sucesso no arcabouço da IFT no contexto de segmentação de imagens, permitindo incorporar informações de alto nível, tais como, restrições de forma, polaridade de borda e restrição de conexidade, a fim de customizar a segmentação para um dado objeto desejado. Funções não monotonicamente incrementais foram também exploradas com sucesso na geração de superpixels, via sequências de execuções da IFT. Neste trabalho, apresentamos um estudo sobre a Transformada Imagem-Floresta Diferencial no caso de funções NMI. Nossos estudos indicam que o algoritmo da DIFT original apresenta uma série de inconsistências para funções não monotonicamente incrementais. Este trabalho estende a DIFT, visando incorporar um subconjunto das funções NMI em grafos dirigidos e mostrar sua aplicação no contexto da geração de superpixels. Outra aplicação que é apresentada para difundir a relevância das funções NMI é o algoritmo Bandeirantes para perseguição de bordas e rastreamento de curvas. / Image segmentation is a problem of great relevance in computer vision, in which an image is divided into relevant regions, such as to isolate an object of interest for a given application. Segmentation methods with monotonically incremental connectivity functions (MI) based on the Image Foresting Transform (IFT) have achieved great success in several contexts. In interactive segmentation of images, in which the user is allowed to specify the desired object, new seeds can be added and/or removed to correct the labeling until achieving the expected segmentation. This process generates a sequence of IFTs that can be calculated more efficiently by the Differential Image Foresting Trans- form (DIFT). Recently, non-monotonically incremental connectivity functions (NMI) have been used successfully in the IFT framework in the context of image segmentation, allowing the incorporation of shape, boundary polarity, and connectivity constraints, in order to customize the segmentation for a given target object. Non-monotonically incremental functions were also successfully exploited in the generation of superpixels, via sequences of IFT executions. In this work, we present a study of the Differential Image Foresting Transform in the case of NMI functions. Our research indicates that the original DIFT algorithm presents a series of inconsistencies for non-monotonically incremental functions. This work extends the DIFT algorithm to NMI functions in directed graphs, and shows its application in the context of the generation of superpixels. Another application that is presented to spread the relevance of NMI functions is the Bandeirantes algorithm for curve tracing and boundary tracking.

Nous connaissons depuis plus d’un siècle des bouleversements climatiques, socio-économiques et sociétaux de plus en plus fréquents et intenses. L’adaptation à ce contexte incertain, pour envisager l’avenir de façon durable, est un défi particulièrement important. L’agriculture est au cœur de ce défi et de la réflexion sur les modes d’adaptation. Les anthropologues ont récemment identifié deux types de changements de l’agriculteur: le changement mineur (en accord avec le comportement normatif de référence), et le changement majeur (changement profond, remise en cause des normes majoritaires permettant l’adoption de comportements innovants). Ces deux types de changement sont très proches de l’adaptation incrémentale et de l’adaptation transformationnelle qui sont définis tant pour des individus que pour des institutions locales, nationales ou internationales … Nous proposons un modèle individu-centré de l’adaptation des agriculteurs qui intègre dynamiquement changements mineurs et majeurs, en nous focalisant sur la dimension psycho-sociale de ces changements. Nous appliquons notre modèle à la question, de la conversion à l’agriculture biologique des éleveurs laitiers français. Cette transition a en effet été généralement caractérisée comme un changement majeur, ou une adaptation transformationnelle, et s’avère donc pertinente pour tester notre modèle. Le modèle d’agriculteur est en général dans un régime stable durant lequel seuls des changements mineurs sont opérés. Ces changements sont menés en imitant des comportements adoptés par les agriculteurs les plus crédibles. La crédibilité d’un agriculteur pour un autre est d’autant plus forte que son volume produit est supérieur à celui de cet agriculteur. Ces changements se traduisent concrètement par des modifications des volumes produits qui correspondent à des choix de pratiques plus ou moins intensifs.Le modèle peut passer d’un régime stable à un autre en transitant par un changement majeur impliquant de changer la hiérarchie des importances, accordées aux dimensions « productiviste » et « environnementale », qui pondèrent son évaluation des résultats de son mode de production (par exemple : conventionnel ou biologique). Le calcul de son évaluation utilise la théorie de l’action raisonnée. Il permet à l’agriculteur de déterminer sa satisfaction, à partir de ses résultats précédents, et de ses résultats comparés à ceux de ses pairs crédibles, en tenant compte de l’importance accordées à chaque dimension d’évaluation. Lorsqu’un agriculteur est insatisfait de son mode de production courant, il évalue sa satisfaction potentielle pour un autre mode de production, et adopte ce dernier si son gain de satisfaction est supérieur à un seuil. Il change alors la hiérarchie des importances accordées à ses dimensions d’évaluation pour être en accord avec le mode adopté. Un agriculteur biologique accordera ainsi par exemple une importance plus forte à la dimension environnementale qu’à la dimension productiviste. Ce changement implique alors une réévaluation tant de la crédibilité de ses pairs que de leurs pratiques. (...) / For more than a century, climatic, socio-economic and societal changes are more and more frequent and intense. Adapting to this uncertain context to envisage a sustainable future is a particularly important challenge. Agriculture is at the heart of this challenge and the reflection on the modes of adaptation. Anthropologists have recently identified two types of farmer changes: minor change (consistent with normative behavior), and major change (deep change, challenge the majority norms allowing adoption of innovative behaviors). These two types of change are very close to the incremental adaptation and the transformational adaptation that are defined for individuals as well as for local, national or international institutions.We propose an individual-based model to study farmers’ adaptations that dynamically integrates minor and major changes. We focus on the social-psychological dimension of these changes. Our model is applied to the question of French dairy farmers’ conversions to organic farming. This transition has been characterized as a major change, or a transformational adaptation, and is therefore relevant to test our model.A farmer stays generally in a stable regime doing only minor changes. These changes are carried out by imitating the practices of the most credible farmers. The credibility of one farmer given to another is larger if his (her) produced volume is greater than that of this farmer. These changes are shown by concrete changes of produced volumes, which correspond to more or less intensive choices of practice.In the model, a farmer passes from one stable regime to another through a major change involving a change of his (her) hierarchy of importance over the "productivist" and "environmental" evaluative dimensions. The importance weights his (her) evaluation of the results according to the mode of production (for example: conventional or organic). The computation of the evaluation is based on the theory of reasoned action. This evaluation shows a farmer’s satisfaction that is based on his (her) previous results and his (her) results compared to those of his (her) credible peers, taking into account the importance given to each dimension of evaluation. When a farmer is dissatisfied with his (her) current mode of production, he (she) evaluates his (her) potential satisfaction with another mode of production, and adopts the latter if his (her) satisfaction gain is above a threshold. He (she) then changes the importance hierarchy given to the evaluation dimensions to be in agreement with the adopted mode. For example, an organic farmer will given more importance to the environmental dimension than to the productivist dimension. This change implies a reassessment of both the credibility given to his (her) peers and their practices. (...)