We discuss two topics in this talk. First we study compact Ricci-flat four dimensional manifolds without boundary and obtain point wise restrictions on curvature( not involving global quantities such as volume and diameter) which force the metric to be flat. We obtain the same conclusion for compact Ricci-flat K¨ahler surfaces with similar but weaker restrictions on holomorphic sectional curvature. Next we study the reaction ODE associated to the evolution of the Riemann curvature operator along the Ricci flow. We analyze the behavior of this ODE near algebraic curvature operators of certain special type that includes the Riemann curvature operators of various(locally) symmetric spaces. We explicitly show the existence of some solution curves to the ODE connecting the curvature operators of certain symmetric spaces. Although the results of these two themes are different, the underlying common feature is the reaction ODE which plays an important role in both.

L'IRM de diffusion (IRMd) est une modalité d'imagerie qui permet d'étudier in vivo la structure des faisceaux de la substance blanche grâce à la caractérisation des propriétés de diffusion des molécules d'eau dans le cerveau. Les travaux de cette thèse se sont concentrés sur la comparaison de groupes d'individus en IRMd. Le but est d'identifier les zones de la substance blanche dont les propriétés structurelles sont statistiquement différentes entre les deux populations ou significativement corrélées avec certaines variables explicatives. L’enjeu est de pouvoir localiser et caractériser les lésions causées par une pathologie et de comprendre les mécanismes sous-jacents. Pour ce faire, nous avons proposé dans cette thèse des méthodes d'analyse basées voxel reposant sur le Modèle Linéaire Général (MLG) et ses extensions multivariées et sur des variétés, qui permettent d'effectuer des tests statistiques intégrant explicitement des variables explicatives. En IRMd, la diffusion des molécules d'eau peut être modélisée par un tenseur d'ordre deux représenté par une matrice symétrique définie-positive de dimension trois. La principale contribution de cette thèse a été de montrer la plus-value de considérer, dans le MLG, l'information complète du tenseur par rapport à un unique descripteur scalaire caractérisant la diffusion (fraction d’anisotropie ou diffusion moyenne), comme cela est généralement fait dans les études en neuro-imagerie. Plusieurs stratégies d’extension du MLG aux tenseurs ont été comparées, que ce soit en termes d’hypothèse statistique (homoscédasticité vs hétéroscédasticité), de métrique utilisée pour l’estimation des paramètres (Euclidienne, Log-Euclidienne et Riemannienne), ou de prise en compte de l’information du voisinage spatial. Nous avons également étudié l'influence de certains prétraitements comme le filtrage et le recalage. Enfin, nous avons proposé une méthode de caractérisation des zones détectées afin d’en faciliter l’interprétation physiopathologique. Les validations ont été menées sur données synthétiques ainsi que sur une base d’images issues d’une cohorte de patients atteints de Neuromyélite optique de Devic. / Diffusion Tensor MRI (DT-MRI) is an imaging modality that allows to study in vivo the structure of white matter fibers through the characterization of diffusion properties of water molecules in the brain. This work focused on group comparison in DT-MRI. The aim is to identify white matter regions whose structural properties are statistically different between two populations or significantly correlated with some explanatory variables. The challenge is to locate and characterize lesions caused by a disease and to understand the underlying mechanisms. To this end, we proposed several voxel-based strategies that rely on the General Linear Model (GLM) and its multivariate and manifold-based extensions, to perform statistical tests that explicitly incorporate explanatory variables. In DT-MRI, diffusion of water molecules can be modeled by a second order tensor represented by a three dimensional symmetric and positive definite matrix. The main contribution of this thesis was to demonstrate the added value of considering the full tensor information as compared to a single scalar index characterizing some diffusion properties (fractional anisotropy or mean diffusion) in the GLM, as it is usually done in neuroimaging studies. Several strategies for extending the GLM to tensor were compared, either in terms of statistical hypothesis (homoscedasticity vs heteroscedasticity), or metrics used for parameter estimation (Euclidean, Log-Euclidean and Riemannian), or the way to take into account the spatial neighborhood information. We also studied the influence of some pre-processing such as filtering and registration. Finally, we proposed a method for characterizing the detected regions in order to facilitate their physiopathological interpretation. Validations have been conducted on synthetic data as well as on a cohort of patients suffering from Neuromyelitis Optica.

Almost contact metric submersions constitute a class of Riemannian submersions whose total space is an almost contact metric manifold. Regarding the base space, two types are studied. Submersions of type I are those whose base space is an almost contact metric manifold while, when the base space is an almost Hermitian manifold, then the submersion is said to be of type II. After recalling the known notions and fundamental properties to be used in the sequel, relationships between the structure of the fibres with that of the total space are established. When the fibres are almost Hermitian manifolds, which occur in the case of a type I submersions, we determine the classes of submersions whose fibres are Kählerian, almost Kählerian, nearly Kählerian, quasi Kählerian, locally conformal (almost) Kählerian, Gi-manifolds and so on. This can be viewed as a classification of submersions of type I based upon the structure of the fibres. Concerning the fibres of a type II submersions, which are almost contact metric manifolds, we discuss how they inherit the structure of the total space. Considering the curvature property on the total space, we determine its corresponding on the fibres in the case of a type I submersions. For instance, the cosymplectic curvature property on the total space corresponds to the Kähler identity on the fibres. Similar results are obtained for Sasakian and Kenmotsu curvature properties. After producing the classes of submersions with minimal, superminimal or umbilical fibres, their impacts on the total or the base space are established. The minimality of the fibres facilitates the transference of the structure from the total to the base space. Similarly, the superminimality of the fibres facilitates the transference of the structure from the base to the total space. Also, it is shown to be a way to study the integrability of the horizontal distribution. Totally contact umbilicity of the fibres leads to the asymptotic directions on the total space. Submersions of contact CR-submanifolds of quasi-K-cosymplectic and quasi-Kenmotsu manifolds are studied. Certain distributions of the under consideration submersions induce the CR-product on the total space. / Mathematical Sciences / D. Phil. (Mathematics)

Recognition of complex events in consumer uploaded Internet videos, captured under realworld settings, has emerged as a challenging area of research across both computer vision and multimedia community. In this dissertation, we present a systematic decomposition of complex events into hierarchical components and make an in-depth analysis of how existing research are being used to cater to various levels of this hierarchy and identify three key stages where we make novel contributions, keeping complex events in focus. These are listed as follows: (a) Extraction of novel semi-global features – firstly, we introduce a Lie-algebra based representation of dominant camera motion present while capturing videos and show how this can be used as a complementary feature for video analysis. Secondly, we propose compact clip level descriptors of a video based on covariance of appearance and motion features which we further use in a sparse coding framework to recognize realistic actions and gestures. (b) Construction of intermediate representations – We propose an efficient probabilistic representation from low-level features computed from videos, based on Maximum Likelihood Estimates which demonstrates state of the art performance in large scale visual concept detection, and finally, (c) Modeling temporal interactions between intermediate concepts – Using block Hankel matrices and harmonic analysis of slowly evolving Linear Dynamical Systems, we propose two new discriminative feature spaces for complex event recognition and demonstrate significantly improved recognition rates over previously proposed approaches.

L'objet de cette thèse est l'étude, sur les variétés riemanniennes compactes $(V_n,g)$ de dimension $n>4$, de l'équation aux dérivées partielles elliptique de quatrième ordre $$(E)\; \Delta^2u+\nabla [a(x)\nabla u] +h(x)u= f(x)|u|^(N-2)u$$ où $a$, $h$, $f$ sont fonction $C^\infty $, avec $f(x)$ fonction constante ou partout positive et $N=(2n\over((n-4)))$ est l'exposant critique. En utilisant la méthode variationnelle on prouve dans le théorème principal que l'équation $(E)$ admet une solution $C^((5,\alpha))(V)$ $0<\alpha<1$ non nulle si une certaine condition qui dépend de la meilleure constante dans les inclusion de Sobolev ($H_2\subset L_(2n\over(n-4))$) est satisfaite. De plus on montre que si $a$ et $h$ sont des fonctions constantes bien précisées la solution de l'équation est positive et $C^\infty(V)$. Lorsque $n\geq 6$, on donne aussi des applications du théorème principal. Dans la dernière partie de cette thèse sur une variété riemannienne compacte à bord de dimension $n$, $(\overline(W)_n,g )$ nous nous intéressons au problème : $$ (P_N) \; \left\lbrace \begin(array)(c) \Delta^2 v+\nabla [a(x)\nabla u] +h(x) v= f(x)|v |^(N-2)v \; \hbox(sur)\; W \\ \Delta v =\delta \, , \, v = \eta \;\hbox(sur) \;\partial W \end(array)\right.$$ avec $\delta$,$\eta$,$f$ fonctions $C^\infty (\overline (W))$ avec $f(x)$ fonction partout positive et on démontre l'existence d'une solution non triviale pour le problème $(P_N)$.

No description available.

Lors des dernières décennies, la Séparation Aveugle de Sources (BSS) est devenue un outil de premier plan pour le traitement de données multi-valuées. L’objectif de ce doctorat est cependant d’étudier les cas grande échelle, pour lesquels la plupart des algorithmes classiques obtiennent des performances dégradées. Ce document s’articule en quatre parties, traitant chacune un aspect du problème: i) l’introduction d’algorithmes robustes de BSS parcimonieuse ne nécessitant qu’un seul lancement (malgré un choix d’hyper-paramètres délicat) et fortement étayés mathématiquement; ii) la proposition d’une méthode permettant de maintenir une haute qualité de séparation malgré un nombre de sources important: iii) la modification d’un algorithme classique de BSS parcimonieuse pour l’application sur des données de grandes tailles; et iv) une extension au problème de BSS parcimonieuse non-linéaire. Les méthodes proposées ont été amplement testées, tant sur données simulées que réalistes, pour démontrer leur qualité. Des interprétations détaillées des résultats sont proposées. / During the last decades, Blind Source Separation (BSS) has become a key analysis tool to study multi-valued data. The objective of this thesis is however to focus on large-scale settings, for which most classical algorithms fail. More specifically, it is subdivided into four sub-problems taking their roots around the large-scale sparse BSS issue: i) introduce a mathematically sound robust sparse BSS algorithm which does not require any relaunch (despite a difficult hyper-parameter choice); ii) introduce a method being able to maintain high quality separations even when a large-number of sources needs to be estimated; iii) make a classical sparse BSS algorithm scalable to large-scale datasets; and iv) an extension to the non-linear sparse BSS problem. The methods we propose are extensively tested on both simulated and realistic experiments to demonstrate their quality. In-depth interpretations of the results are proposed.

We consider the problem of numerical integration, where one aims to approximate an integral of a given continuous function from the function values at given sampling points, also known as quadrature points. A useful framework for such an approximation process is provided by the theory of reproducing kernel Hilbert spaces and the concept of the worst case quadrature error. However, the computation of optimal quadrature points, which minimize the worst case quadrature error, is in general a challenging task and requires efficient algorithms, in particular for large numbers of points. The focus of this thesis is on the efficient computation of optimal quadrature points on the torus T^d, the sphere S^d, and the rotation group SO(3). For that reason we present a general framework for the minimization of the worst case quadrature error on Riemannian manifolds, in order to construct numerically such quadrature points. Therefore, we consider, for N quadrature points on a manifold M, the worst case quadrature error as a function defined on the product manifold M^N. For the optimization on such high dimensional manifolds we make use of the method of steepest descent, the Newton method, and the conjugate gradient method, where we propose two efficient evaluation approaches for the worst case quadrature error and its derivatives. The first evaluation approach follows ideas from computational physics, where we interpret the quadrature error as a pairwise potential energy. These ideas allow us to reduce for certain instances the complexity of the evaluations from O(M^2) to O(M log(M)). For the second evaluation approach we express the worst case quadrature error in Fourier domain. This enables us to utilize the nonequispaced fast Fourier transforms for the torus T^d, the sphere S^2, and the rotation group SO(3), which reduce the computational complexity of the worst case quadrature error for polynomial spaces with degree N from O(N^k M) to O(N^k log^2(N) + M), where k is the dimension of the corresponding manifold. For the usual choice N^k ~ M we achieve the complexity O(M log^2(M)) instead of O(M^2). In conjunction with the proposed conjugate gradient method on Riemannian manifolds we arrive at a particular efficient optimization approach for the computation of optimal quadrature points on the torus T^d, the sphere S^d, and the rotation group SO(3). Finally, with the proposed optimization methods we are able to provide new lists with quadrature formulas for high polynomial degrees N on the sphere S^2, and the rotation group SO(3). Further applications of the proposed optimization framework are found due to the interesting connections between worst case quadrature errors, discrepancies and potential energies. Especially, discrepancies provide us with an intuitive notion for describing the uniformity of point distributions and are of particular importance for high dimensional integration in quasi-Monte Carlo methods. A generalized form of uniform point distributions arises in applications of image processing and computer graphics, where one is concerned with the problem of distributing points in an optimal way accordingly to a prescribed density function. We will show that such problems can be naturally described by the notion of discrepancy, and thus fit perfectly into the proposed framework. A typical application is halftoning of images, where nonuniform distributions of black dots create the illusion of gray toned images. We will see that the proposed optimization methods compete with state-of-the-art halftoning methods.

We consider the problem of numerical integration, where one aims to approximate an integral of a given continuous function from the function values at given sampling points, also known as quadrature points. A useful framework for such an approximation process is provided by the theory of reproducing kernel Hilbert spaces and the concept of the worst case quadrature error. However, the computation of optimal quadrature points, which minimize the worst case quadrature error, is in general a challenging task and requires efficient algorithms, in particular for large numbers of points. The focus of this thesis is on the efficient computation of optimal quadrature points on the torus T^d, the sphere S^d, and the rotation group SO(3). For that reason we present a general framework for the minimization of the worst case quadrature error on Riemannian manifolds, in order to construct numerically such quadrature points. Therefore, we consider, for N quadrature points on a manifold M, the worst case quadrature error as a function defined on the product manifold M^N. For the optimization on such high dimensional manifolds we make use of the method of steepest descent, the Newton method, and the conjugate gradient method, where we propose two efficient evaluation approaches for the worst case quadrature error and its derivatives. The first evaluation approach follows ideas from computational physics, where we interpret the quadrature error as a pairwise potential energy. These ideas allow us to reduce for certain instances the complexity of the evaluations from O(M^2) to O(M log(M)). For the second evaluation approach we express the worst case quadrature error in Fourier domain. This enables us to utilize the nonequispaced fast Fourier transforms for the torus T^d, the sphere S^2, and the rotation group SO(3), which reduce the computational complexity of the worst case quadrature error for polynomial spaces with degree N from O(N^k M) to O(N^k log^2(N) + M), where k is the dimension of the corresponding manifold. For the usual choice N^k ~ M we achieve the complexity O(M log^2(M)) instead of O(M^2). In conjunction with the proposed conjugate gradient method on Riemannian manifolds we arrive at a particular efficient optimization approach for the computation of optimal quadrature points on the torus T^d, the sphere S^d, and the rotation group SO(3). Finally, with the proposed optimization methods we are able to provide new lists with quadrature formulas for high polynomial degrees N on the sphere S^2, and the rotation group SO(3). Further applications of the proposed optimization framework are found due to the interesting connections between worst case quadrature errors, discrepancies and potential energies. Especially, discrepancies provide us with an intuitive notion for describing the uniformity of point distributions and are of particular importance for high dimensional integration in quasi-Monte Carlo methods. A generalized form of uniform point distributions arises in applications of image processing and computer graphics, where one is concerned with the problem of distributing points in an optimal way accordingly to a prescribed density function. We will show that such problems can be naturally described by the notion of discrepancy, and thus fit perfectly into the proposed framework. A typical application is halftoning of images, where nonuniform distributions of black dots create the illusion of gray toned images. We will see that the proposed optimization methods compete with state-of-the-art halftoning methods.