Submitted by Alex Melges Barbosa (alex.melgesb@gmail.com) on 2017-03-02T15:41:01Z No. of bitstreams: 1 PDF_final_cd - Alex.pdf: 1031527 bytes, checksum: d71a279fd3f7c761ebeb48e598e07b22 (MD5) / Approved for entry into archive by LUIZA DE MENEZES ROMANETTO (luizamenezes@reitoria.unesp.br) on 2017-03-08T13:52:00Z (GMT) No. of bitstreams: 1 barbosa_am_me_sjrp.pdf: 1031527 bytes, checksum: d71a279fd3f7c761ebeb48e598e07b22 (MD5) / Made available in DSpace on 2017-03-08T13:52:00Z (GMT). No. of bitstreams: 1 barbosa_am_me_sjrp.pdf: 1031527 bytes, checksum: d71a279fd3f7c761ebeb48e598e07b22 (MD5) Previous issue date: 2017-02-22 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Neste trabalho, apresentaremos uma descrição axiomática das classes de Stiefel-Whitney e, assumindo válidos estes axiomas, mostraremos algumas de suas aplicações. Posteriormente, definiremos as classes de Stiefel-Whitney e mostraremos que esta definição satisfaz os axiomas, além de garantir a unicidade das classes de Stiefel-Whitney. Por fim, definiremos a classe de Euler e mostraremos algumas de suas aplicações, bem como sua relação com as classes de Stiefel-Whitney. / In this work, we will present an axiomatic description of the Stiefel-Whitney classes and, taking these axioms true, we will show some of their applications. After that, we will define the Stiefel-Whitney classes and we will show this definition meets the axioms, besides it ensures the unity of the Stiefel-Whitney classes. Lastly, we will define the Euler class and we will show some of its applications as well as its relationship with the Stiefel-Whitney classes.

Σκοπός της εργασίας μας είναι η μελέτη κάποιων αναγωγικών χώρων που παρουσιάζουν ενδιαφέρουσα γεωμετρία. Συγκεκριμένα, μελετάμε τη γεωμετρία της σφαίρας S^n όταν αυτή είναι αμφιδιαφορική με έναν χώρο πηλίκο G/K και την γεωμετρία των πολλαπλοτήτων Stiefel SO(n)/SO(n-k) (το σύνολο όλων των k-πλαισίων του R^n). Ένας ομογενής χώρος αποτελεί επέκταση των ομάδων Lie, καθώς είναι μια λεία πολλαπλότητα M στην οποία δρα μεταβατικά μια ομάδα Lie G. Κάθε τέτοιος χώρος δίνεται ως M = G/K, όπου K = {g\in G : gp = p} (p \in M). Η βασική γεωμετρική ιδιότητα των ομογενών χώρων είναι ότι αν γνωρίζουμε την τιμή κάποιου γεωμετρικού μεγέθους σε ένα σημείο του χώρου, τότε μπορούμε να υπολογίσουμε την τιμή του μεγέθους αυτού σε οποιοδήποτε άλλο σημείο. Το ιδιαίτερο χαρακτηριστικό των αναγωγικών χώρων G/K είναι ότι υπάρχει ένας Ad(K)-αναλλοίωτος υπόχωρος της άλγεβρας Lie(G). Η περιγραφή όλων των μεταβατικών δράσεων μιας ομάδας Lie σε μια πολλαπλότητα M αποτελεί ένα δύσκολο πρόβλημα. Για την περίπτωση των σφαιρών αυτές έχουν περιγραφτεί το 1953 από τους Montgomery-Samelson-Borel. Στην εργασία μας μελετάμε τη γεωμετρία (καμπυλότητες, μετρικές Einstein) των σφαιρών S^3, S^5 όταν αυτές είναι αμφιδιαφορικές με τα πηλίκα S^3 = SO(4)/SO(3) = SU(2) και S^5 = SO(6)/SO(5) = SU(3)/SU(2). Αντίστοιχα προβλήματα εξετάζονται για τις πολλαπλότητες Stiefel SO(n)/SO(n-k), όπου η περιγραφή όλων των SO(n)-αναλλοίωτων μετρικών παρουσιάζει δυσκολία, λόγω του ότι η ισοτροπική αναπαράστασή τους περιέχει ισοδύναμα υποπρότυπα. Μελετάμε για ποιές από τις συγκεκριμένες πολλαπλότητες η μετρική που επάγεται από τη μορφή Killing είναι μετρική Einstein και περιγράφουμε αναλυτικά τις διαγώνιες SO(n)-αναλλοίωτες μετρικές Einstein στις πολλαπλότητες SO(n)/SO(n-2). Επιπλέον παρουσιάζουμε και ένα καινούργιο αποτέλεσμα, ότι στην πολλαπλότητα SO(5)/SO(2) οι μοναδικές SO(5)-αναλλοίωτες μετρικές Einstein είναι οι μετρικές που είχαν βρεθεί από τον Jensen το 1973. / The purpose of our work is to study homogeneous spaces that present interesting geometry. These include the geometry of the sphere S^n diffeomorphic to a quotient space G/K and the geometry of Stiefel manifolds SO(n)/SO(n-k) (the set of all k-planes in R^n). A homogeneous space is a smooth manifold M in which a Lie group acts transitively. Any such space is given as M = G/K where K = {g\in G : gp = p} (p\in M). The basic geometric property of homogeneous space is that if we know the value of a geometrical object at a point of the space, then we can estimate the value of thiw quantity at any other point. The special feature of reductive homogeneous space G/K is that there exists an Ad(K)-invariant subspace of the Lie algebra Lie(G). The description of all transitive actions of a Lie group into a manifold M is a difficult problem. In the case of spheres such actions have been described in 1953 by the Montgomery, Samelson and Borel. In our work we study the geometry (curvature, Einstein metrics) of the sphere S^3 = SO(4)/SO(3) = SU(2), S^5 = SO(6)/SO(5) = SU(3)/SU(2). Similar problems are examined for the Stiefel manifolds SO(n)/SO(n-k). The description of all SO(n)-invariant metrics presents serious difficulties because the isotropy representation contains equivalent submodules. We study for which of the manifolds SO(n)/SO(n-k) the metric induced by the Killing form is an Einstein metric and we describe in detail the diagonal SO(n)-invariant Einstein metrics on the Stiefel manifolds SO(n)/SO(n-2). In addition, we give the new result that for the Stiefel manifold SO(5)/SO(2) the unique SO(5)-invariant Einstein metrics are the metrics found by Jensen in 1973.

Made available in DSpace on 2016-06-02T20:27:39Z (GMT). No. of bitstreams: 1 4150.pdf: 1194001 bytes, checksum: afed3d1198e69a86efd9f7b69562b509 (MD5) Previous issue date: 2012-03-05 / Universidade Federal de Minas Gerais / Let (Mm; T) be a smooth involution on a closed smooth m-dimensional manifold and F = n [j=0 Fj (n < m) its fixed point set, where Fj denotes the union of those components of F having dimension j. The famous Five Halves Theorem of J. Boardman, announced in 1967, establishes that, if F is nonbounding, then m _ 5 2n; further, this estimative is best possible. In this work, we obtain improvements of this theorem, by imposing certain conditions on F. The main result of the work is in Chapter 4, where the improvements in question are obtained by taking into account the decomposability degree of the components of F. Specifically, let ! = (i1; i2; :::; it) be a non-dyadic partition of j, 2 _ j _ n, and s!(x1; x2; :::; xj) the smallest symmetric polynomial over Z2 on degree one variables x1; x2; :::; xj containing the monomial xi1 1 xi2 2 :::xit t . Write s!(Fj) 2 Hj(Fj ;Z2) for the usual cohomology class corresponding to s!(x1; x2; :::; xj). The decomposability degree of Fj , denoted by l(Fj), is the minimum length of a non-dyadic partition ! with s!(Fj) 6= 0 (here, the length of ! = (i1; i2; :::; it) is t). Suppose the fixed point set of (Mm; T) has the form F = ( j [k=0 Fk) [ Fn, where 2 _ j < n < m and Fj is nonbounding. Write n &#56256;&#56320; j = 2pq, where q _ 1 is odd and p _ 0, and set m(n &#56256;&#56320; j) = 2n + p &#56256;&#56320; q + 1 if p _ q and m(n&#56256;&#56320;j) = 2n+2p&#56256;&#56320;q if p _ q. Then we prove that m _ m(n&#56256;&#56320;j)+2j +l(Fj). In addition, given a non-dyadic partition ! = (i1; i2; : : : ; it) of j, 2 _ j < n, we develop a method to construct involutions (Mm; T) with F of the form F = ([k<j Fk)[Fj[Fn, where m = m(n &#56256;&#56320; j) + 2j + t and s![Fj ] 6= 0, for special values of n; j and !. In some special cases, this method shows that the above bound is best possible. For example, this gives the following improvement of the Five Halves Theorem: if the fixed point set F = n [j=0 Fj of (Mm; T) has Fn&#56256;&#56320;1 and Fn nonbounding, then m _ minf2n + l(Fn&#56256;&#56320;1); 2n + l(Fn)g; further, the bounds m _ 2n + l(Fn&#56256;&#56320;1) and m _ 2n + l(Fn) are separately best possible. Other consequence: if the fixed point set F = n [j=0 Fj of (Mm; T) has n = 2k, k _ 3 and vii Fn&#56256;&#56320;1 nonbounding, then m _ 5k &#56256;&#56320; 2, and this bound is best possible (the Five Halves Theorem says that m _ 5k). We also deal with the low codimension phenomenon, which is expressed by the fact that for certain F the codimension m &#56256;&#56320; n is too small; here, the advances obtained are concerned with the fact that, in the considered cases, the number of components of F is not limited as a function of n (in the literature one finds results of this nature with F having two, three or four components). For example, among the results obtained one has: if F has the form F = F3 [ ( n [j=0 j even Fj), with n _ 4 even, and all involved normal bundles are nonbounding, then m _ n + 4; further, this estimative is best possible. Finally, we also study bounds for the case F = Fn [ F4, considering that in the literature one has results involving F = Fn [ Fi for i = 0; 1; 2; 3. For example, we show that if the fixed set of (Mm; T) has the form F = Fn [ F4, n is odd and the normal bundle over F4 is not a boundary, then m _ n + 5; further, this bound is best possible. / Sejam (Mm; T) uma involução suave em uma variedade m-dimensional, fechada e suave Mm e F = n [j=0 Fj (n < m) o seu conjunto de pontos fixos, onde Fj denota a união das componentes de F com dimensão j. O famoso 5=2-Teorema de J. Boardman, anunciado em 1967, estabelece que, se F é não bordante, então m _ 5 2n; além disso, esta estimativa é a melhor possível. Neste trabalho, nós obtemos melhorias para este teorema, impondo certas condições sobre F. O resultado principal se encontra no Capítulo 4, onde as melhorias em questão são obtidas levando-se em conta o grau de decomponibilidade das componentes de F. Especificamente, seja ! = (i1; i2; :::; it) uma partição não diádica de j, 2 _ j _ n, e seja s!(x1; x2; :::; xj) a menor polinomial simétrica sobre Z2, nas variáveis de grau um x1; x2; :::; xj , contendo o monômio xi1 1 xi2 2 :::xit t . Escreva s!(Fj) 2 Hj(Fj ;Z2) para a classe usual de cohomologia correspondente a s!(x1; x2; :::; xj). O grau de decomponibilidade de Fj , denotado por l(Fj), é o menor comprimento de uma partição não diádica ! com s!(Fj) 6= 0 (aqui, o comprimento de ! = (i1; i2; :::; it) é t). Suponhamos que o conjunto de pontos fixos de (Mm; T) tem a forma F = ( j [k=0 Fk) [ Fn, onde 2 _ j < n < m e Fj é não bordante. Escreva n &#56256;&#56320; j = 2pq, onde q _ 1 é ímpar e p _ 0, e tome m(n&#56256;&#56320;j) = 2n+p&#56256;&#56320;q+1, se p _ q, e m(n&#56256;&#56320;j) = 2n+2p&#56256;&#56320;q, se p _ q. Então, provamos que m _ m(n&#56256;&#56320;j)+2j +l(Fj). Em adição, dada uma partição não diádica ! = (i1; i2; : : : ; it) de j, 2 _ j < n, desenvolvemos um método para construir involuções (Mm; T) com F da forma F = ([k<j Fk) [ Fj [ Fn, onde m = m(n &#56256;&#56320; j) + 2j + t e s![Fj ] 6= 0, para valores especiais de n, j e !. Em alguns casos específicos, este método mostra que o limitante acima é o melhor possível. Por exemplo, tal método fornece a seguinte melhoria para o 5=2-Teorema de J. Boardman: se o conjunto de pontos fixos F = n [j=0 Fj de (Mm; T) possui Fn&#56256;&#56320;1 e Fn não bordantes, então m _ minf2n+l(Fn&#56256;&#56320;1); 2n+l(Fn)g; além disso, os limitantes m _ 2n + l(Fn&#56256;&#56320;1) e m _ 2n + l(Fn) são separadamente os melhores possíveis. Outra consequência: se o conjunto de pontos fixos F = n [j=0 Fj de (Mm; T) tem n = 2k, k _ 3 e Fn&#56256;&#56320;1 não bordante, então m _ 5k &#56256;&#56320; 2, e este limitante é o melhor possível (o 5=2-Teorema diz que m _ 5k, nesse caso). Nós também trabalhamos com alguns casos envolvendo fenômenos de baixa codimensão, caracterizados pelo fato que, para específicos conjuntos de pontos fixos F, a codimensão m &#56256;&#56320; n é muito pequena; aqui, os avanços obtidos nos casos considerados relacionam-se à circunstância do número de componentes de F não ser limitado como uma função de n (na literatura, encontramos resultados dessa natureza onde F possui 2, 3 ou 4 componentes). Como exemplo dos resultados obtidos, temos o seguinte: se F tem a forma F = F3 [( n [j=0 j par Fj), com n _ 4 par, e tal que todos os fibrados normais envolvidos são não bordantes, então m _ n + 4; além disso, esta estimativa é a melhor possível. Finalmente, trabalhamos com limitantes para o caso F = Fn [F4, considerandose que na literatura atual temos alguns resultados envolvendo F = Fn [ Fi, para i = 0; 1; 2; 3. Por exemplo, nós mostramos que se o conjunto de pontos fixos de (Mm; T) tem a forma F = Fn [ F4, com n ímpar, e o fibrado normal sobre F4 é não bordante, então m _ n + 5; além disso, esse limitante é o melhor possível.

Made available in DSpace on 2016-06-02T20:28:27Z (GMT). No. of bitstreams: 1 4588.pdf: 701327 bytes, checksum: 07aaf91b8be59a3db7c6c5cf38e55c59 (MD5) Previous issue date: 2012-09-24 / Financiadora de Estudos e Projetos / We present an introductory study of smooth manifolds, bundles and Stiefel- Whitney classes (of real vector bundles). We explained that, given a certain smooth m-dimensional manifold, the Stiefel- Whitney classes of its tangent bundle can be used to ensure that such a manifold does not immerse (smoothly) in certain Euclidean spaces Rj . In this sense, we consider the Grassmann manifold G2;n of the 2-subspaces of Rn+2, and we carry out a detailed study of the following non-immersion theorem, proved by V. Oproiu [Proceedings of the Edinburgh Mathematical Society, 1977]: "Let n > 1 be a natural number and consider s = 2r such that s _ 2n < 2s. If n = s - 1, then G2;n does not immerse in R2s-3; if n = s - 1, then G2;n does not immerse in R3s-3." / Apresentamos um estudo introdutório de Variedades Suaves, Fibrados e Classes de Stiefel-Whitney (de _brados vetorias reais). Explicamos que, dada uma certa variedade suave m-dimensional, as classes de Stiefel-Whitney do seu _brado tangente podem ser usadas para garantir que tal variedade não imerge (suavemente) em certos espaços Euclidianos Rj . Nesse sentido, consideramos a variedade Grassmanniana G2;n, variedade dos 2-subespaços de Rn+2, e realizamos um estudo detalhado do seguinte teorema de não imersão, provado por V. Oproiu [Proceedings of the Edinburgh Mathematical Society, 1977]: "Seja n > 1 um natural e considere s = 2r tal que s _ 2n < 2s. Se n 6= s &#1048576; 1, então G2;n não imerge em R2s-3; se n = s - 1, então G2;n não imerge em R3s-3."

Made available in DSpace on 2016-06-02T20:28:27Z (GMT). No. of bitstreams: 1 4588.pdf: 701327 bytes, checksum: 07aaf91b8be59a3db7c6c5cf38e55c59 (MD5) Previous issue date: 2012-09-24 / Financiadora de Estudos e Projetos / We present an introductory study of smooth manifolds, bundles and Stiefel- Whitney classes (of real vector bundles). We explained that, given a certain smooth m-dimensional manifold, the Stiefel- Whitney classes of its tangent bundle can be used to ensure that such a manifold does not immerse (smoothly) in certain Euclidean spaces Rj . In this sense, we consider the Grassmann manifold G2;n of the 2-subspaces of Rn+2, and we carry out a detailed study of the following non-immersion theorem, proved by V. Oproiu [Proceedings of the Edinburgh Mathematical Society, 1977]: "Let n > 1 be a natural number and consider s = 2r such that s < ou = 2n < 2s. If n different s - 1, then G2;n does not immerse in R2s-3; if n = s - 1, then G2;n does not immerse in R3s-3." / Apresentamos um estudo introdutório de Variedades Suaves, Fibrados e Classes de Stiefel-Whitney (de fibrados vetorias reais). Explicamos que, dada uma certa variedade suave m-dimensional, as classes de Stiefel-Whitney do seu fibrado tangente podem ser usadas para garantir que tal variedade não imerge (suavemente) em certos espaços Euclidianos Rj . Nesse sentido, consideramos a variedade Grassmanniana G2;n, variedade dos 2-subespaços de Rn+2, e realizamos um estudo detalhado do seguinte teorema de não imersão, provado por V. Oproiu [Proceedings of the Edinburgh Mathematical Society, 1977]: "Seja n > 1 um natural e considere s = 2r tal que s < ou = 2n < 2s. Se n for diferente de s - 1, então G2;n não imerge em R2s-3; se n = s - 1, então G2;n não imerge em R3s-3."

The problem of embedding a manifold in Euclidean space is considered. Manifolds are introduced in Chapter I along with other basic definitions and examples. Chapter II contains a proof of the Regular Value Theorem along with the "Easy" Whitney Embedding Theorem. In Chapter III, vector bundles are introduced and some of their properties are discussed. Chapter IV introduces the Stiefel-Whitney classes and the four properties that characterize them. Finally, in Chapter V, the Stiefel-Whitney classes are used to produce a lower bound on the dimension of Euclidean space that is needed to embed real projective space.

Cette thèse généralise au problème spatial dans le cas lunaire les études sur diverses familles de mouvements quasi-périodiques dans le problème plan des trois corps. En tronquant au premier ordre non trivial le développement en puissances du rapport des demi grands axes de la fonction perturbatrice moyennée sur les angles rapides, on obtient un système complètement intégrable qui peut servir de première approximation pour le système initial. C'est le système quadripolaire, découvert par Harrington. Dans un article classique, Lidov et Ziglin ont étudié la dynamique de ce système. Nous commençons par établir l'existence de solutions quasi-périodiques du problème spatial des trois corps en appliquant les théorèmes de KAM à ce système. Nous montrons ensuite l'existence de familles de solutions que nous appelons solutions quasi-périodiques de quasi-collision : ce sont des solutions le long desquelles deux des corps deviennent arbitrairement proches l'un de l'autre sans toutefois avoir de collision : la limite inférieure de leur distance est nulle alors que la limite supérieure est strictement positive. Ces solutions sont quasi-périodiques dans un système régularisé à un changement de temps près. Des solutions de ce type ont été mises en évidence tout d'abord dans le problème restreint plan circulaire par Chenciner et Llibre puis, dans le problème plan des trois corps par Féjoz. Nous prouvons l'existence d'une mesure positive de ces solutions dans le problème spatial des trois corps. L'existence de ce type de solutions avait été prédit par Marchal dont nous confirmons rigoureusement le résultat. La démonstration consiste en l'application d'un théorème KAM équivariant dans une régularisation du problème, ici celle de Kustaanheimo-Stiefel, et par la compréhension, suivant Féjoz, de la relation entre régularisation et moyennisation.

In many areas of science one often has a given matrix, representing for example a measured data set and is required to find a matrix that is closest in a suitable norm to the matrix and possesses additionally a structure, inherited from the model used or coming from the application. We call these problems structured matrix nearness problems. We look at three different groups of these problems that come from real applications, analyze the properties of the corresponding matrix structure, and propose algorithms to solve them efficiently. The first part of this thesis concerns the nearness problem of finding the nearest k factor correlation matrix C(X) = diag(I_n -XX T)+XX T to a given symmetric matrix, subject to natural nonlinear constraints on the elements of the n x k matrix X, where distance is measured in the Frobenius norm. Such problems arise, for example, when one is investigating factor models of collateralized debt obligations (CDOs) or multivariate time series. We examine several algorithms for solving the nearness problem that differ in whether or not they can take account of the nonlinear constraints and in their convergence properties. Our numerical experiments show that the performance of the methods depends strongly on the problem, but that, among our tested methods, the spectral projected gradient method is the clear winner. In the second part we look at two two-sided optimization problems where the matrix of unknowns Y ε R {n x p} lies in the Stiefel manifold. These two problems come from an application in atomic chemistry where one is looking for atomic orbitals with prescribed occupation numbers. We analyze these two problems, propose an analytic optimal solution of the first and show that an optimal solution of the second problem can be found by solving a convex quadratic programming problem with box constraints and p unknowns. We prove that the latter problem can be solved by the active-set method in at most 2p iterations. Subsequently, we analyze the set of optimal solutions C}= {Y ε R n x p:Y TY=I_p,Y TNY=D} of the first problem for N symmetric and D diagonal and find that a slight modification of it is a Riemannian manifold. We derive the geometric objects required to make an optimization over this manifold possible. We propose an augmented Lagrangian-based algorithm that uses these geometric tools and allows us to optimize an arbitrary smooth function over C. This algorithm can be used to select a particular solution out of the latter set C by posing a new optimization problem. We compare it numerically with a similar algorithm that ,however, does not apply these geometric tools and find that our algorithm yields better performance. The third part is devoted to low rank nearness problems in the Q-norm, where the matrix of interest is additionally of linear structure, meaning it lies in the set spanned by s predefined matrices U₁,..., U_s ε {0,1} n x p. These problems are often associated with model reduction, for example in speech encoding, filter design, or latent semantic indexing. We investigate three approaches that support any linear structure and examine further the geometric reformulation by Schuermans et al. (2003). We improve their algorithm in terms of reliability by applying the augmented Lagrangian method and show in our numerical tests that the resulting algorithm yields better performance than other existing methods.

Cette thèse étudie deux modèles paramétriques et non paramétriques pour le changement de représentation. L'objectif des deux modèles diffère. Le premier cherche une représentation en plus grande dimension pour gagner en robustesse. L'objectif est de répartir uniformément l’information d’un signal sur toutes les composantes de sa représentation en plus grande dimension. La recherche d'un tel code s'exprime comme un problème inverse impliquant une régularisation de type norme infinie. Nous proposons une formulation bayésienne du problème impliquant une nouvelle loi de probabilité baptisée démocratique, qui pénalise les fortes amplitudes. Deux algorithmes MCMC proximaux sont présentés pour approcher des estimateurs bayésiens. La méthode non supervisée présentée est appelée BAC-1. Des expériences numériques illustrent les performances de l’approche pour la réduction de facteur de crête. Le second modèle identifie un sous-espace pertinent de dimension réduite à des fins de modélisation. Mais les méthodes probabilistes proposées nécessitent généralement de fixer à l'avance la dimension du sous-espace. Ce travail introduit BNP-PCA, une version bayésienne non paramétrique de l'analyse en composantes principales. La méthode couple une loi uniforme sur les bases orthonormales à un a priori non paramétrique de type buffet indien pour favoriser une utilisation parcimonieuse des composantes principales et aucun réglage n'est nécessaire. L'inférence est réalisée à l'aide des méthodes MCMC. L'estimation de la dimension du sous-espace et le comportement numérique de BNP-PCA sont étudiés. Nous montrons la flexibilité de BNP-PCA sur deux applications / This thesis proposes Bayesian parametric and nonparametric models for signal representation. The first model infers a higher dimensional representation of a signal for sake of robustness by enforcing the information to be spread uniformly. These so called anti-sparse representations are obtained by solving a linear inverse problem with an infinite-norm penalty. We propose in this thesis a Bayesian formulation of anti-sparse coding involving a new probability distribution, referred to as the democratic prior. A Gibbs and two proximal samplers are proposed to approximate Bayesian estimators. The algorithm is called BAC-1. Simulations on synthetic data illustrate the performances of the two proposed samplers and the results are compared with state-of-the art methods. The second model identifies a lower dimensional representation of a signal for modelisation and model selection. Principal component analysis is very popular to perform dimension reduction. The selection of the number of significant components is essential but often based on some practical heuristics depending on the application. Few works have proposed a probabilistic approach to infer the number of significant components. We propose a Bayesian nonparametric principal component analysis called BNP-PCA. The proposed model involves an Indian buffet process to promote a parsimonious use of principal components, which is assigned a prior distribution defined on the manifold of orthonormal basis. Inference is done using MCMC methods. The estimators of the latent dimension are theoretically and empirically studied. The relevance of the approach is assessed on two applications

Nous présentons un nouvel algorithme d’apprentissage de circuits quantiques basé sur la descente de gradient classique. Comme ce sujet unifie deux disciplines, nous expliquons les deux domaines aux gens de l’autre discipline. Conséquemment, nous débutons par une présentation du calcul quantique et des circuits quantiques pour les gens en apprentissage automatique suivi d’une présentation des algorithmes d’apprentissage automatique pour les gens en informatique quantique. Puis, pour motiver et mettre en contexte nos résultats, nous passons à une légère revue de littérature en apprentissage automatique quantique. Ensuite, nous présentons notre modèle, son algorithme, ses variantes et quelques résultats empiriques. Finalement, nous critiquons notre implémentation en montrant des extensions et des nouvelles approches possibles. Les résultats principaux se situent dans ces deux dernières parties, qui sont respectivement les chapitres 4 et 5 de ce mémoire. Le code de l’algorithme et des expériences que nous avons créé pour ce mémoire se trouve sur notre github à l’adresse suivante : https://github.com/AldoLamarre/quantumcircuitlearning. / We present a new learning algorithm for quantum circuits based on gradient descent. Since this subject unifies two areas of research, we explain each field for people working in the other domain. Consequently, we begin by introducing quantum computing and quantum circuits to machine learning specialists, followed by an introduction of machine learning to quantum computing specialists. To give context and motivate our results we then give a light literature review on quantum machine learning. After this, we present our model, its algorithms and its variants, then discuss our currently achieved empirical results. Finally, we criticize our models by giving extensions and future work directions. These last two parts are our main results. They can be found in chapter 4 and 5 respectively. Our code which helped obtain these results can be found on github at this link : https://github.com/ AldoLamarre/quantumcircuitlearning.