[pt] Os sistemas dinâmicos desempenham um papel crucial no que diz respeito à compreensão de fenômenos inerentes a diversos campos da ciência. Desde a última década, todo aporte tecnológico alcançado ao longo de anos de investigação deram origem a uma estratégia orientada a dados, permitindo a inferência de modelos capazes de representar sistemas dinâmicos. Além disso, independentemente dos tipos de sensores adotados a fim de realizar o procedimento de aquisição de dados, é natural verificar a existência de uma certa corrupção ruidosa nos referidos dados. Genericamente, a tarefa de identificação é diretamente afetada pelo cenário ruidoso previamente descrito, implicando na falsa descoberta de um modelo generalizável. Em outras palavras, a corrupção ao ruído pode ser responsável pela geração de uma representação matemática infiel de um determinado sistema. Nesta tese, no que diz respeito à tarefa de identificação, é demonstrado como a robustez ao ruído pode ser melhorada a partir da hibridização de técnicas de aprendizado de máquina, como aumento de dados, regressão esparsa, seleção de características, extração de características, critério de informação, pesquisa em grade e validação cruzada. Especificamente, sob as perspectivas de classificação e regressão, o sucesso da estratégia proposta é apresentado a partir de exemplos numéricos, como o crescimento logístico, oscilador Duffing, modelo FitzHugh-Nagumo, atrator de Lorenz e uma modelagem Suscetível-Infeccioso-Recuperado (SIR) do Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2). / [en] The dynamical systems play a fundamental role related to the understanding of the phenomena inherent to several fields of science. Since the last decade, all technological advances achieved throughout years of research have given rise to a data oriented strategy, enabling the inference of dynamical systems. Moreover, regardless the sensor types adopted to perform the data acquisition procedure, it is natural to verify the existence of a certain noise corruption in such data. Generically, the identification task is directly affected by the noisy scenario previously described, which entails in the false discovery of a generalizable model. In other words, the noise corruption might be responsible to give rise to a worthless mathematical representation of a given system. In this thesis, with respect to the identification assignment, it is demonstrated how the robustness to noise may be improved from the hybridization of machine learning techniques, such as data augmentation, sparse regression, feature selection, feature extraction, information criteria, grid search and cross validation. Specifically, through classification and regression perspectives, the success of the proposed strategy is presented from numerical examples, such as the logistic growth, Duffing oscillator, FitzHugh–Nagumo model, Lorenz attractor and a Susceptible-Infectious-Recovered (SIR) modeling of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2).

No description available.

No description available.

Dans cette thèse, nous nous intéressons à l’étude des courbes discrètes bruitées qui correspondent aux contours d’objets dans des images. Nous avons proposé plusieurs outils permettant de les analyser. Les points dominants (points dont l’estimation de la courbure est localement maximale) jouent un rôle très important dans la reconnaissance de formes et, nous avons développé une méthode non heuristique, rapide et fiable pour les détecter dans une courbe discrète. Cette méthode est une amélioration d’une méthode existante introduite par Nguyen et al. La nouvelle méthode consiste à calculer une mesure d’angle. Nous avons proposé aussi deux approches pour la simplification polygonale : une méthode automatique minimisant, et une autre fixant le nombre de sommets du polygone résultant. Ensuite, nous avons introduit un nouvel outil géométrique, nommé couverture tangentielle adaptative (ATC), reposant sur la détection des épaisseurs significatives introduites par Kerautret et al. Ces épaisseurs calculées en chaque point du contour à analyser, permettent d’estimer localement le niveau de bruit. Dans ce contexte notre algorithme de construction de la couverture tangentielle adaptative prend en considération les différents niveaux de bruits présents dans la courbe à étudier et ne nécessite pas de paramètre. Deux applications de l’ATC sont proposées en analyse d’images : d’une part la décomposition des contours d’une forme dans une image en arcs et en segments de droite et d’autre part, dans le cadre d’un projet avec une université d’Inde, autour du langage des signes et la reconnaissance des gestes de la main. Premièrement, la méthode de décomposition des courbes discrètes en arcs et en segments de droite est basée sur deux outils : la détection de points dominants en utilisant la couverture tangentielle adaptative et la représentation dans l’espace des tangentes du polygone, issue des points dominants détectés. Les expériences montrent la robustesse de la méthode w.r.t. le bruit. Deuxièmement, à partir des contours des mains extraits d’images prises par une Kinect, nous proposons différents descripteurs reposant sur des points dominants sélectionnés du contour des formes dans les images. Les descripteurs proposés, qui sont une combinaison entre descripteurs statistiques et descripteurs géométriques, sont efficaces et conviennent à la reconnaissance de gestes / In this thesis, we are interested in the study of noisy discrete curves that correspond to the contours of objects in images. We have proposed several tools to analyze them. The dominant points (points whose curvature estimation is locally maximal) play a very important role in pattern recognition and we have developed a non-heuristic, fast and reliable method to detect them in a discrete curve. This method is an improvement of an existing method introduced by Nguyen et al. The new method consists in calculating a measure of angle. We have also proposed two approaches for polygonal simplification: an automatic method minimizing, and another fixing the vertex number of the resulting polygon. Then we proposed a new geometric tool, called adaptive tangential cover ATC, based on the detection of meaningful thickness introduced by Kerautret et al. These thicknesses are calculated at each point of the contours allow to locally estimate the noise level. In this context our construction algorithm of adaptive tangential cover takes into account the different levels of noise present in the curve to be studied and does not require a parameter. Two applications of ATC in image analysis are proposed: on the one hand the decomposition of the contours of a shape in an image into arcs and right segments and on the other hand, within the framework of a project with an Indian university about the sign language and recognition of hand gestures. Firstly, the method to decompose discrete curves into arcs and straight segments is based on two tools: dominant point detection using adaptive tangential cover and tangent space representation of the polygon issued from detected dominant points. The experiments demonstrate the robustness of the method w.r.t. noise. Secondly, from the outlines of the hands extracted from images taken by a Kinect, we propose several descriptors from the selected dominant points computed from the adaptive tangential cover. The proposed descriptors, which are a combination of statistical descriptors and geometrical descriptors, are effective and suitable for gesture recognition

<p>The problem under consideration is an unconstrained mini-mization problem in noisy environment. The common approach for solving the problem is Stochastic Approximation (SA) algorithm. We propose a class of adaptive step size schemes for the SA algorithm. The step size selection in the proposed schemes is based on the objective functionvalues. At each iterate, interval estimates of the optimal function&nbsp; value are constructed using the xed number of previously observed function values.&nbsp;If the observed function value in the current iterate is larger than the upper bound of the interval, we reject the current iterate. If the observed function value in the current iterate is smaller than the lower bound of the interval, we suggest a larger step size in the next iterate. Otherwise, if the function value lies in the interval, we propose a small safe step size in the next iterate. In this manner, a faster progress of the algorithm is ensured&nbsp;when it is expected that larger steps will improve the performance of the algorithm. We propose two main schemes which dier in the intervals that we construct at each iterate. In the rst scheme, we construct a symmetrical interval that can be viewed as a condence-like interval for the optimal function value. The bounds of the interval are shifted means of the xed number of previously observed function values. The generalization&nbsp;of this scheme using a convex combination instead of the mean is also presented. In the second scheme, we use the minimum and the maximum of previous noisy function values as the lower and upper bounds of the interval, respectively. The step size sequences generated by the proposed schemes satisfy the step size convergence conditions for the SA algorithm almost surely. Performance of SA algorithms with the new step size schemes is tested on a set of standard test problems. Numerical results&nbsp;support theoretical expectations and verify eciency of the algorithms in comparison to other relevant modications of SA algorithms. Application of the algorithms in LASSO regression models is also considered. The algorithms are applied for estimation of the regression parameters where the objective function contains L₁ penalty.</p> / <p>Predmet istraživanja doktorske disertacije su numerički postupci za re&scaron;avanje problema stohastičke optimizacije. Najpoznatiji numerički postupak za re&scaron;avanje pomenutog problema je algoritam stohastičke aproksimacije (SA). U disertaciji se&nbsp;&nbsp; predlaže nova klasa &scaron;ema za prilagođavanje dužina koraka u svakoj iteraciji. Odabir dužina koraka u predloženim &scaron;emama se zasniva na vrednostima funkcije cilja. U svakoj iteraciji formira se intervalna ocena optimalne vrednosti funkcije cilja koristeći samo registrovane vrednosti funkcije cilja iz ksnog broja prethodnih iteracija. Ukoliko je vrednost funkcije cilja u trenutnoj iteraciji veća od gornje granice intervala, iteracija se odbacuje. Korak dužine 0 se koristi u narednoj iteraciji. Ako je trenutna vrednost funkcije cilja manja od donje granice intervala, predlaže se duži korak u narednoj iteraciji. Ukoliko vrednost funkcije leži u intervalu, u narednoj iteraciji se koristi korak dobijen harmonijskim pravilom. Na ovaj način se obezbeđuje brzi progres algoritma i&nbsp; izbegavaju mali koraci posebno kada se povećava broj iteracija.&nbsp; &Scaron;eme izbegavaju korake proporcionalne sa 1/k kada se očekuje da ce duži koraci pobolj&scaron;ati proces optimizacije. Predložene &scaron;eme se razlikuju u intervalima koji se formiraju u svakoj iteraciji. U prvoj predloženoj &scaron;emi se formira ve&scaron;tački interval poverenja za ocenu optimalne vrednosti funkcije cilja u svakoj iteraciji. Granice tog intervala se uzimaju za&nbsp; kriterijume dovoljnog smanjenja ili rasta funkcije cilja. Predlaže se i uop&scaron;tenje ove &scaron;eme tako &scaron;to se umesto srednje vrednosti koristi konveksna kombinacija prethodnih vrednosti funkcije cilja. U drugoj &scaron;emi, kriterijum po kom se prilagođavaju dužine koraka su minimum i maksimum prethodnih registrovanih vrednosti funkcije cilja. Nizovi koji se formiranju predloženim &scaron;emama zadovoljavaju uslove potrebne za konvergenciju SA algoritma skoro sigurno. SA algoritmi sa novim &scaron;emama za prilagođavanje dužina koraka su testirani na standardnim test&nbsp; problemima i upoređ eni sa SA algoritmom i njegovim postojećim modikacijama. Rezultati pokazuju napredak u odnosu na klasičan algoritam stohastičke aproksimacije sa determinističkim nizom dužine koraka kao i postojećim adaptivnim algoritmima. Takođe se razmatra primena novih algoritama na LASSO regresijske modele. Algoritmi su primenjeni za ocenjivanje parametara modela.</p>

No description available.

The research described in this thesis was motivated by the need of a robust model capable of representing 3D data obtained with 3D sensors, which are inherently noisy. In addition, time constraints have to be considered as these sensors are capable of providing a 3D data stream in real time. This thesis proposed the use of Self-Organizing Maps (SOMs) as a 3D representation model. In particular, we proposed the use of the Growing Neural Gas (GNG) network, which has been successfully used for clustering, pattern recognition and topology representation of multi-dimensional data. Until now, Self-Organizing Maps have been primarily computed offline and their application in 3D data has mainly focused on free noise models, without considering time constraints. It is proposed a hardware implementation leveraging the computing power of modern GPUs, which takes advantage of a new paradigm coined as General-Purpose Computing on Graphics Processing Units (GPGPU). The proposed methods were applied to different problem and applications in the area of computer vision such as the recognition and localization of objects, visual surveillance or 3D reconstruction.

Nowadays, new computers generation provides a high performance that enables to build computationally expensive computer vision applications applied to mobile robotics. Building a map of the environment is a common task of a robot and is an essential part to allow the robots to move through these environments. Traditionally, mobile robots used a combination of several sensors from different technologies. Lasers, sonars and contact sensors have been typically used in any mobile robotic architecture, however color cameras are an important sensor due to we want the robots to use the same information that humans to sense and move through the different environments. Color cameras are cheap and flexible but a lot of work need to be done to give robots enough visual understanding of the scenes. Computer vision algorithms are computational complex problems but nowadays robots have access to different and powerful architectures that can be used for mobile robotics purposes. The advent of low-cost RGB-D sensors like Microsoft Kinect which provide 3D colored point clouds at high frame rates made the computer vision even more relevant in the mobile robotics field. The combination of visual and 3D data allows the systems to use both computer vision and 3D processing and therefore to be aware of more details of the surrounding environment. The research described in this thesis was motivated by the need of scene mapping. Being aware of the surrounding environment is a key feature in many mobile robotics applications from simple robotic navigation to complex surveillance applications. In addition, the acquisition of a 3D model of the scenes is useful in many areas as video games scene modeling where well-known places are reconstructed and added to game systems or advertising where once you get the 3D model of one room the system can add furniture pieces using augmented reality techniques. In this thesis we perform an experimental study of the state-of-the-art registration methods to find which one fits better to our scene mapping purposes. Different methods are tested and analyzed on different scene distributions of visual and geometry appearance. In addition, this thesis proposes two methods for 3d data compression and representation of 3D maps. Our 3D representation proposal is based on the use of Growing Neural Gas (GNG) method. This Self-Organizing Maps (SOMs) has been successfully used for clustering, pattern recognition and topology representation of various kind of data. Until now, Self-Organizing Maps have been primarily computed offline and their application in 3D data has mainly focused on free noise models without considering time constraints. Self-organising neural models have the ability to provide a good representation of the input space. In particular, the Growing Neural Gas (GNG) is a suitable model because of its flexibility, rapid adaptation and excellent quality of representation. However, this type of learning is time consuming, specially for high-dimensional input data. Since real applications often work under time constraints, it is necessary to adapt the learning process in order to complete it in a predefined time. This thesis proposes a hardware implementation leveraging the computing power of modern GPUs which takes advantage of a new paradigm coined as General-Purpose Computing on Graphics Processing Units (GPGPU). Our proposed geometrical 3D compression method seeks to reduce the 3D information using plane detection as basic structure to compress the data. This is due to our target environments are man-made and therefore there are a lot of points that belong to a plane surface. Our proposed method is able to get good compression results in those man-made scenarios. The detected and compressed planes can be also used in other applications as surface reconstruction or plane-based registration algorithms. Finally, we have also demonstrated the goodness of the GPU technologies getting a high performance implementation of a CAD/CAM common technique called Virtual Digitizing.

Konversationerna i ett kundforum rör sig över olika ämnen och språket är inkonsekvent. Texterna uppfyller inte de krav som brukar ställas på material inför automatisk nyckelordsextraktion. Uppsatsens undersöker hur nyckelord automatiskt kan extraheras ur ett kundforum trots dessa svårigheter. Fokus i undersökningen ligger på tre aspekter av nyckelordsextraktion. Den första faktorn rör hur den etablerade nyckelordsextraktionsmetoden TF*IDF presterar jämfört med fyra metoder som skapas med hänsyn till materialets ovanliga struktur. Nästa faktor som testas är om olika sätt att räkna ordfrekvens påverkar resultatet. Den tredje faktorn är hur metoderna presterar om de endast använder inläggen, rubrikerna eller båda texttyperna i sina extraktioner. Icke-parametriska test användes för utvärdering av extraktionerna. Ett antal Friedmans test visar att metoderna i några fall skiljer sig åt gällande förmåga att identifiera relevanta nyckelord. I post-hoc-test mellan de högst presterande metoderna ses en av de nya metoderna i ett fall prestera signifikant bättre än de andra nya metoderna men inte bättre än TF*IDF. Ingen skillnad hittades mellan användning av olika texttyper eller sätt att räkna ordfrekvens. För framtida forskning rekommenderas reliabilitetstest av manuellt annoterade nyckelord. Ett större stickprov bör användas än det i aktuell studie och olika förslag ges för att förbättra rättning av extraherade nyckelord. / Conversations in a customer forum span across different topics and the language is inconsistent. The text type do not meet the demands for automatic keyword extraction. This essay examines how keywords can be automatically extracted despite these difficulties. Focus in the study are three areas of keyword extraction. The first factor regards how the established keyword extraction method TF*IDF performs compared to four methods created with the unusual material in mind. The next factor deals with different ways to calculate word frequency. The third factor regards if the methods use only posts, only titles, or both in their extractions. Non-parametric tests were conducted to evaluate the extractions. A number of Friedman's tests shows the methods in some cases differ in their ability to identify relevant keywords. In post-hoc tests performed between the highest performing methods, one of the new methods perform significantly better than the other new methods but not better than TF*IDF. No difference was found between the use of different text types or ways to calculate word frequency. For future research reliability test of manually annotated keywords is recommended. A larger sample size should be used than in the current study and further suggestions are given to improve the results of keyword extractions.

Nous nous proposons d'étudier la gestion du bruit stochastique d'expression génique. On s'intéresse plus particulièrement à la dynamique du bruit lors de la réponse cellulaire. Comment évolue le bruit? Quels sont les mécanismes en jeux? Quelle est l'importance du bruit dans le fonctionnement cellulaire? Pour répondre à ces questions, nous nous appuyons sur le réseau de régulation génétique qui gère la réponse au stress nutritionnel chez E. Coli. L'étude du comportement dynamique de ce réseau, au niveau d'une population de bactéries, a été initiée et est portée par la forte collaboration de deux équipes de la région : une de bio-informaticiens (l'équipe de Hidde de Jong de l'INRIA Rhône-Alpes) et la deuxième de biologistes (l'équipe de Hans Geiselmann, Laboratoire d'Adaptation et Pathogénie des Micro-organismes). En profitant donc de l'expérience et de la compréhension acquise par ces équipes, nous étudions les réponses individuelles de chaque bactérie lors de la transition entre état de stress nutritionnel, et état de croissance exponentielle. Le bruit d'expression génique est quantifié dans des nœuds clés du réseau de régulation. Pour ce faire, les bactéries sont suivies individuellement par microscopie de fluorescence sur plusieurs générations. Les données de fluorescence collectées sur cellules uniques permettent d'étudier la variabilité inter-cellulaire. Cette variabilité est quantifiée tout le long de la réponse: à chaque instant, on connaît la distribution des densités de fluorescence cellulaire dans la population de cellules. Et le suivi des lignées individuelles permet de travailler sur une population de cellules saines: les individus malades ou morts qui ne se divisent pas, sont écartés. En réduisant ainsi les phénomènes cellulaires en jeux, on réduit le nombre de paramètres. Les sources de bruit sont moins nombreuses, et il est plus facile de comprendre les mécanismes en jeux. Les informations de lignage cellulaire permettent aussi d'étudier la variabilité introduite par la phase du cycle cellulaire: les événements de division cellulaire peut être artificiellement synchronisés. Le bruit est alors étudié sur une population en phase lors de la division. Cette étude montre que le bruit sondé n'est pas dominé par les différences dans la phase du cycle cellulaire. On peut donc modéliser nos cellules sans tenir compte des différences introduites par le cycle cellulaire. Le modèle testé est simplifié aux étapes de transcription-traduction-maturation. Les paramètres du modèle sont inférés de nos données expérimentales, et le modèle est testé à travers des simulations. / We aim to investigate the management of the stochastic noise in gene expression and more precisely the study of noise in dynamical cellular responses. How the noise varies following a perturbation? What mechanisms are at play? How important is noise in the cellular function? To answer these questions, we are interested in the genetic regulatory network that handles the nutritional stress response in E. Coli. The noise of gene expression is quantified in a key node of the network control. For that bacteria are followed individually by fluorescence and phase contrast microscopy over several generations. This variability between cells is quantified throughout the response to the nutritional perturbation: at every moment, we know the density distribution of cellular fluorescence in the cell population. And monitoring of individual lines allows us to take into account only the population of healthy cells: individuals that do not divide neither grow, are discarded. Thereby reducing other sources of variability (e.g. cellular phenomena) we reduce the number of parameters. Noise sources are less numerous, and it is easier to understand the mechanisms at play. Also the information on cell lineage allow to study the variability introduced by the phase of the cell cycle: the events of cell division can be artificially synchronized. This study shows that the noise sounded is not dominated by differences in the phase of the cell cycle. We can therefore model our cells regardless of the differences introduced by the cell cycle. The tested model is simplified to the steps of transcription-translation-maturation. The model parameters are inferred from our experimental data and the model is tested through simulations.