Cette thèse s’inscrit dans le domaine mathématique de l’analyse des équations aux dérivées partielles non linéaires. Précisément, nous nous sommes intéressés à une classe de problèmes elliptiques et paraboliques avec coefficients singuliers. Ce manque de régularité pose un certain nombre de difficultés qui ne permettent pas d’utiliser directement les méthodes classiques de l’analyse non-linéaire fondées entre autres sur des résultats de compacité. Dans les démonstrations des principaux résultats, nous montrons comment pallier ces difficultés. Ceci suppose d’adapter certaines techniques bien connues mais aussi d’introduire de nouvelles méthodes. Dans ce contexte, une étape importante est l’estimation fine du comportement des solutions qui permet d’adapter le principe de comparaison faible, d’utiliser la régularité elliptique et parabolique et d’appliquer dans un nouveau contexte la théorie globale de la bifurcation analytique. La thèse se présente sous forme de deux parties indépendantes. 1- Dans la première partie (chapitre I de la thèse), nous avons étudié un problème quasi-linéaire parabolique fortement singulier faisant intervenir l’opérateur p-Laplacien. On a démontré l’existence locale et la régularité de solutions faibles. Ce résultat repose sur des estimations a priori obtenues via l’utilisation d’inégalités de type log-Sobolev combinées à des inégalités de Gagliardo-Nirenberg. On démontre l’unicité de la solution pour un intervalle de valeurs du paramètre de la singularité en utilisant un principe de comparaison faible fondé sur la monotonie d’un opérateur non linéaire adéquat. 2- Dans la deuxième partie (correspondant aux Chapitres II, III et IV de la thèse), nous sommes intéressés à l’étude de problèmes de bifurcation globale. On a établi pour ces problèmes l’existence de continuas non bornés de solutions qui admettent localement une paramétrisation analytique. Pour établir ces résultats, nous faisons appel à différents outils d’analyse non linéaire. Un outil important est la théorie analytique de la bifurcation globale qui a été introduite par Dancer (voir Chapitre II de la thèse). Pour un problème semi linéaire elliptique avec croissance critique en dimension 2, on montre que les solutions le long de la branche convergent vers une solution singulière (solution non bornée) lorsque la norme des solutions converge vers l’infini. Par ailleurs nous montrons que la branche admet une infinité dénombrable de "points de retournement" correspondant à un changement de l’indice de Morse des solutions qui tend vers l’infini le long de la branche. / This thesis is concerned with the mathematical study of nonlinear partial differential equations. Precisely, we have investigated a class of nonlinear elliptic and parabolic problems with singular coefficients. This lack of regularity involves some difficulties which prevent the straight-orward application of classical methods of nonlinear analysis based on compactness results. In the proofs of the main results, we show how to overcome these difficulties. Precisely we adapt some well-known techniques together with the use of new methods. In this framework, an important step is to estimate accurately the solutions in order to apply the weak comparison principle, to use the regularity theory of parabolic and elliptic equations and to develop in a new context the analytic theory of global bifurcation. The thesis presents two independent parts. 1- In the first part (corresponding to Chapter I), we are interested by a nonlinear and singular parabolic equation involving the p-Laplacian operator. We established for this problem that for any non-negative initial datum chosen in a certain Lebeque space, there exists a local positive weak solution. For that we use some a priori bounds based on logarithmic Sobolev inequalities to get ultracontractivity of the associated semi-group. Additionaly, for a range of values of the singular coefficient, we prove the uniqueness of the solution and further regularity results. 2- In the second part (corresponding to Chapters II, III and IV of the thesis), we are concerned with the study of global bifurcation problems involving singular nonlinearities. We establish the existence of a piecewise analytic global path of solutions to these problems. For that we use crucially the analytic bifurcation theory introduced by Dancer (described in Chapter II of the thesis). In the frame of a class of semilinear elliptic problems involving a critical nonlinearity in two dimensions, we further prove that the piecewise analytic path of solutions admits asymptotically a singular solution (i.e. an unbounded solution), whose Morse index is infinite. As a consequence, this path admits a countable infinitely many “turning points” where the Morse index is increasing.

Uma imagem engloba informação que precisa ser organizada para interpretar e compreender seu conteúdo. Existem diversas técnicas computacionais para extrair a principal informação de uma imagem e podem ser divididas em três áreas: análise de cor, textura e forma. Uma das principais delas é a análise de forma, por descrever características de objetos baseadas em seus pontos fronteira. Propomos um método de caracterização de imagens, por meio da análise de forma, baseada nas propriedades espectrais do laplaciano em grafos. O procedimento construiu grafos G baseados nos pontos fronteira do objeto, cujas conexões entre vértices são determinadas por limiares T_l. A partir dos grafos obtêm-se a matriz de adjacência A e a matriz de graus D, as quais definem a matriz Laplaciana L=D -A. A decomposição espectral da matriz Laplaciana (autovalores) é investigada para descrever características das imagens. Duas abordagens são consideradas: a) Análise do vetor característico baseado em limiares e a histogramas, considera dois parâmetros o intervalo de classes IC_l e o limiar T_l; b) Análise do vetor característico baseado em vários limiares para autovalores fixos; os quais representam o segundo e último autovalor da matriz L. As técnicas foram testada em três coleções de imagens: sintéticas (Genéricas), parasitas intestinais (SADPI) e folhas de plantas (CNShape), cada uma destas com suas próprias características e desafios. Na avaliação dos resultados, empregamos o modelo de classificação support vector machine (SVM), o qual avalia nossas abordagens, determinando o índice de separação das categorias. A primeira abordagem obteve um acerto de 90 % com a coleção de imagens Genéricas, 88 % na coleção SADPI, e 72 % na coleção CNShape. Na segunda abordagem, obtém-se uma taxa de acerto de 97 % com a coleção de imagens Genéricas; 83 % para SADPI e 86 % no CNShape. Os resultados mostram que a classificação de imagens a partir do espectro do Laplaciano, consegue categorizá-las satisfatoriamente. / An image includes information that needs to be organized to interpret and understand its contents. There are several computational techniques to extract the main information of images and are divided into three areas: color, texture and shape analysis. One of the main of them is shape analysis, since it describes objects getting main features based on reference points, usually border points. This dissertation proposes a shape analysis method based on the spectral properties of the Laplacian in graphs to represent images. The procedure builds G graphs based on object border points, whose connections between vertices are determined by thresholds T_l. From graphs G we obtain the adjacency matrix A and matrix degrees D, which define the Laplacian matrix L=D -A. Thus, spectral decomposition of the Laplacian matrix (eigenvalues) is investigated to describe image features. Two approaches are considered: a)Analysis of feature vector based on thresholds and histograms, it considers two parameters, classes range IC_l and threshold T_l; b) Analysis of feature vector based on multiple linear for fixed eigenvalues, which represents the second and final eigenvalue matrix L. The techniques were tested in three image datasets: synthetic (Generic), human intestinal parasites (SADPI) and plant leaves (CNShape), each of these with its own features and challenges. Afterwards to evaluate our results, we used the classification model Support Vector Machine (SVM) to evaluate our approaches, determining the percentage of separation of categories. The first approach achieved 90 % of precision with the Generic image dataset, 88 % in SADPI dataset, and 72 % in CNShape dataset. In the second approach, it obtains 97 % of precision with the Generic image dataset, 83 % for SADPI and 86 % in CNShape respectively. The results show that the classification of images from the Laplacian spectrum can categorize them satisfactorily.

Conselho Nacional de Desenvolvimento CientÃfico e TecnolÃgico / CoordenaÃÃo de AperfeiÃoamento de Pessoal de NÃvel Superior / Neste trabalho, provaremos alguns resultados sobre espectro essencial de uma classe de variedades Riemannianas, nÃo necessariamente completas, com condiÃÃes de curvatura na vizinhanÃa de um raio. Sobre essas condiÃÃes obtemos que o espectro essencial do operador de Laplace contÃm um intervalo. Como aplicaÃÃo, obteremos o espectro do operador de Laplace de regiÃes ilimitadas dos espaÃos formas, tais como a horobola do espaÃo hiperbÃlico e cones do espaÃo Euclidiano. Construiremos tambÃm um exemplo que indica a necessidade das condiÃÃes globais sobre o supremo das curvaturas seccionais fora de uma bola para que a variedade nÃo tenha espectro essencial. / In this thesis we consider a family of Riemannian manifolds, not necessarily complete, with curvature conditions in a neighborhood of a ray. Under these conditions we obtain that the essential spectrum of the Laplace operator contains an interval. The results presented in this thesis allow to determine the spectrum of the Laplace operator on unlimited regions of space forms, such as horoball in hyperbolic space and cones in Euclidean space. Also construct an example that shows the need of global conditions on the supreme sectional curvature outside a ball, so that the variety has no essential spectrum.

Spectral clustering is often used to partition a data set into a specified number of clusters. Both the unweighted and the vertex-weighted approaches use eigenvectors of the Laplacian matrix of a graph. Our focus is on using vertex-weighted methods to refine clustering of observations. An eigenvector corresponding with the second smallest eigenvalue of the Laplacian matrix of a graph is called a Fiedler vector. Coefficients of a Fiedler vector are used to partition vertices of a given graph into two clusters. A vertex of a graph is classified as unassociated if the Fiedler coefficient of the vertex is close to zero compared to the largest Fiedler coefficient of the graph. We propose a vertex-weighted spectral clustering algorithm which incorporates a vector of weights for each vertex of a given graph to form a vertex-weighted graph. The proposed algorithm predicts association of equidistant or nearly equidistant data points from both clusters while the unweighted clustering does not provide association. Finally, we implemented both the unweighted and the vertex-weighted spectral clustering algorithms on several data sets to show that the proposed algorithm works in general.

Le verre feuilleté est constitué de deux ou plusieurs feuilles de verre assemblées par une ou plusieurs couches de polymères viscoélastiques intermédiaires qui sont à l’origine des interactions entre les mécanismes de cisaillement, de flexion locale et de flexion globale qui régissent le comportement d’ensemble du verre feuilleté. Dans ces travaux, nous proposons un modèle de plaques dites "fortement contrastées" intégrant ces différents mécanismes et leur couplage. Dès lors qu’il existe une séparation d’échelle entre l’épaisseur de la plaque et la taille caractéristique des variations des champs cinématiques, la méthode des développements asymptotiques est applicable pour l’étude des stratifiées. Le comportement effectif de la plaque est dérivé de la loi de comportement 3D des matériaux, associée à une formulation de développements asymptotiques et à une mise à l’échelle appropriée du contraste de rigidité entre les couches raides de verre et des couches souples de polymère viscoélastique. Le procédé fournit une formulation explicite et cohérente, intégrant les efforts duaux des descripteurs macroscopiques (équations constitutives), les équations d’équilibres hors-plan et dans le plan et les différentes lois de comportements du stratifié. Le modèle est d’abord construit dans le cas d’une plaque de verre feuilleté à 2 couches raides puis différents cas de chargements statiques et dynamiques sont appliqués. Il est ensuite étendu aux plaques à 3 couches raides puis généralisé aux plaques à N couches raides. Des campagnes expérimentales sont mises en place pour valider les modèles obtenus. Enfin un cas d’étude de dimensionnement de plancher en verre feuilleté est proposé. / Widely used in recent years, the glass material makes it possible to erect light, aesthetic and resistant structures, in particular with glass lamination method, which consists in assembling two or more sheets of glass by one or more intermediate viscoelastic polymers layer. These viscoelastic interface layers are the source of interactions between the shear, the local bending and the global bending mechanisms that govern the overall behavior of laminated glass. For an optimal dimensioning and an accurate determination of the different fields, we propose a "highlycontrasted" plates model encapsulating these different mechanisms and their coupling. As long as a scale separation between the plate thickness and the caracteristic size of kinematics field variations exists, the asymptotic expansion method can be applied for the study of laminates. The effective plate behaviour is derived from the 3D constitutive law of the materials combined with an asymptotic expansion formulation and an appropriate scaling the tiffness contrast between stiff glass layers and soft viscoelastic polymer layers. The method provides a synthetic and consistent formulation, integrating the dual efforts of the macroscopic descriptors, in-plane and out-of-plane balance equations and the constitutive laws of the laminate. First the model is developped for a 2 stiff layers laminated glass plates, then different static and dynamic loading cases are applied. It is then extended to plates with 3 stiff layers and then generalized to plates with N stiff layers. Experimental campaigns are set up in order to validate the models. Finally, a case study of laminated glass floor design is proposed.

Dans cette thèse, nous menons à l'aide de la méthode des équations intégrales des analyses de sensitivité de solutions ou de spectres de l'équation de conductivité par rapport aux variations géométriques ou de paramètres de l'équation. En particulier, nous considérons le problème de conductivité dans des milieux à forts contrastes, le problème de perturbation du bord d'une inclusion de conductivité, le problème de valeurs propres du Laplacien dans des domaines perturbés et le problème d'ouverture de gap dans le spectre des cristaux photoniques.

<p>In 1989 Jun Kigami made an analytic construction of a Laplacian on the Sierpiński gasket, a construction that he extended to post critically finite fractals. Since then, this field has evolved into a proper theory of analysis on fractals. The new results obtained in this thesis are all in the setting of Kigami's theory. They are presented in three papers.</p><p>Strichartz recently showed that there are first order linear differential equations, based on the Laplacian, that are not solvable on the Sierpiński gasket. In the first paper we give a characterization on the polynomial p so that the differential equation p(Δ)u=f is solvable on any open subset of the Sierpiński gasket for any f continuous on that subset. For general p we find the open subsets on which p(Δ)u=f is solvable for any continuous f.</p><p>In the second paper we describe the infinitesimal geometric behavior of a large class of smooth functions on the Sierpiński gasket in terms of the limit distribution of their local eccentricity, a generalized direction of gradient. The distribution of eccentricities is codified as an infinite dimensional perturbation problem for a suitable iterated function system, which has the limit distribution as an invariant measure. We extend results for harmonic functions found by Öberg, Strichartz and Yingst to larger classes of functions.</p><p>In the third paper we define and study intrinsic first order derivatives on post critically finite fractals and prove differentiability almost everywhere for certain classes of fractals and functions. We apply our results to extend the geography is destiny principle, and also obtain results on the pointwise behavior of local eccentricities. Our main tool is the Furstenberg-Kesten theory of products of random matrices.</p>

This dissertation was to study computational theory and methods for ?nding multiple saddle critical points in Banach spaces. Two local minimax methods were developed for this purpose. One was for unconstrained cases and the other was for constrained cases. First, two local minmax characterization of saddle critical points in Banach spaces were established. Based on these two local minmax characterizations, two local minimax algorithms were designed. Their ?ow charts were presented. Then convergence analysis of the algorithms were carried out. Under certain assumptions, a subsequence convergence and a point-to-set convergence were obtained. Furthermore, a relation between the convergence rates of the functional value sequence and corresponding gradient sequence was derived. Techniques to implement the algorithms were discussed. In numerical experiments, those techniques have been successfully implemented to solve for multiple solutions of several quasilinear elliptic boundary value problems and multiple eigenpairs of the well known nonlinear p-Laplacian operator. Numerical solutions were presented by their pro?les for visualization. Several interesting phenomena of the solutions of quasilinear elliptic boundary value problems and the eigenpairs of the p-Laplacian operator have been observed and are open for further investigation. As a generalization of the above results, nonsmooth critical points were considered for locally Lipschitz continuous functionals. A local minmax characterization of nonsmooth saddle critical points was also established. To establish its version in Banach spaces, a new notion, pseudo-generalized-gradient has to be introduced. Based on the characterization, a local minimax algorithm for ?nding multiple nonsmooth saddle critical points was proposed for further study.

In 1989 Jun Kigami made an analytic construction of a Laplacian on the Sierpiński gasket, a construction that he extended to post critically finite fractals. Since then, this field has evolved into a proper theory of analysis on fractals. The new results obtained in this thesis are all in the setting of Kigami's theory. They are presented in three papers. Strichartz recently showed that there are first order linear differential equations, based on the Laplacian, that are not solvable on the Sierpiński gasket. In the first paper we give a characterization on the polynomial p so that the differential equation p(Δ)u=f is solvable on any open subset of the Sierpiński gasket for any f continuous on that subset. For general p we find the open subsets on which p(Δ)u=f is solvable for any continuous f. In the second paper we describe the infinitesimal geometric behavior of a large class of smooth functions on the Sierpiński gasket in terms of the limit distribution of their local eccentricity, a generalized direction of gradient. The distribution of eccentricities is codified as an infinite dimensional perturbation problem for a suitable iterated function system, which has the limit distribution as an invariant measure. We extend results for harmonic functions found by Öberg, Strichartz and Yingst to larger classes of functions. In the third paper we define and study intrinsic first order derivatives on post critically finite fractals and prove differentiability almost everywhere for certain classes of fractals and functions. We apply our results to extend the geography is destiny principle, and also obtain results on the pointwise behavior of local eccentricities. Our main tool is the Furstenberg-Kesten theory of products of random matrices.

This thesis consists of four chapters. Chapter 1 focuses on theoretical results on high-order laplacian-based regularization in function estimation. We studied the iterated laplacian regularization in the context of supervised learning in order to achieve both nice theoretical properties (like thin-plate splines) and good performance over complex region (like soap film smoother). In Chapter 2, we propose an innovative static path-planning algorithm called m-A* within an environment full of obstacles. Theoretically we show that m-A* reduces the number of vertex. In the simulation study, our approach outperforms A* armed with standard L1 heuristic and stronger ones such as True-Distance heuristics (TDH), yielding faster query time, adequate usage of memory and reasonable preprocessing time. Chapter 3 proposes m-LPA* algorithm which extends the m-A* algorithm in the context of dynamic path-planning and achieves better performance compared to the benchmark: lifelong planning A* (LPA*) in terms of robustness and worst-case computational complexity. Employing the same beamlet graphical structure as m-A*, m-LPA* encodes the information of the environment in a hierarchical, multiscale fashion, and therefore it produces a more robust dynamic path-planning algorithm. Chapter 4 focuses on an approach for the prediction of spot electricity spikes via a combination of boosting and wavelet analysis. Extensive numerical experiments show that our approach improved the prediction accuracy compared to those results of support vector machine, thanks to the fact that the gradient boosting trees method inherits the good properties of decision trees such as robustness to the irrelevant covariates, fast computational capability and good interpretation.