Esta tese examina o problema de detecção de conectividade entre séries temporais no sentido de Granger no caso em que a natureza não linear das interações não permite sua determinação por meio de modelos auto-regressivos lineares vetoriais. Mostra-se que é possível realizar esta detecção com auxílio dos chamados métodos de Kernel, que se tornaram populares em aprendizado por máquina (\'machine learning\') já que tais métodos permitem definir formas generalizadas de teste de Granger, coerência parcial direcionada e função de transferência direcionada. Usando simulações, mostram-se alguns exemplos de detecção nos quais fica também evidente que resultados assintóticos deduzidos originalmente para estimadores lineares podem ser generalizados de modo análogo, mostrando-se válidos no presente contexto kernelizado. / This work examines the connectivity detection problem between time series in the Granger sense when the nonlinear nature of interactions determination is impossible via linear vector autoregressive models, but is, nonetheless, feasible with the aid of the so-called Kernel methods that are popular in machine learning. The kernelization approach allows defining generalised versions for Granger tests, partial directed coherence and directed transfer function, which the simulation of some examples shows that the asymptotic detection results originally deducted for linear estimators, can also be employed under kernelization if suitably adapted.

Esta tese examina o problema de detecção de conectividade entre séries temporais no sentido de Granger no caso em que a natureza não linear das interações não permite sua determinação por meio de modelos auto-regressivos lineares vetoriais. Mostra-se que é possível realizar esta detecção com auxílio dos chamados métodos de Kernel, que se tornaram populares em aprendizado por máquina (\'machine learning\') já que tais métodos permitem definir formas generalizadas de teste de Granger, coerência parcial direcionada e função de transferência direcionada. Usando simulações, mostram-se alguns exemplos de detecção nos quais fica também evidente que resultados assintóticos deduzidos originalmente para estimadores lineares podem ser generalizados de modo análogo, mostrando-se válidos no presente contexto kernelizado. / This work examines the connectivity detection problem between time series in the Granger sense when the nonlinear nature of interactions determination is impossible via linear vector autoregressive models, but is, nonetheless, feasible with the aid of the so-called Kernel methods that are popular in machine learning. The kernelization approach allows defining generalised versions for Granger tests, partial directed coherence and directed transfer function, which the simulation of some examples shows that the asymptotic detection results originally deducted for linear estimators, can also be employed under kernelization if suitably adapted.

Dans ce travail, nous nous intéressons à l’analyse spectrale des systèmes d’opérateurs pseudodifférentiels semi-classiques. Dans la première partie, nous étudions la généralisation du théorème d’Egorov en temps longs dans le cas où l’Hamiltonien quantique qui génère l’évolution en temps et l’observable quantique initiale sont deux opérateurs pseudodifférentiels semiclassiques associés à des symboles à valeurs matricielles. Sous une condition d’hyperbolicité sur le symbole principal de l’Hamiltonien qui assure l’existence des projecteurs semi-classiques, et pour une classe d’observables "semi-classiquement" diagonales par blocs par rapport à ces projecteurs, nous démontrons un théorème de type Egorov valable pour un temps long d’ordre log(h-1) connu comme le temps d’Ehrenfest. Ici h 0 est le paramètre semi-classique. Dans la deuxième partie, nous nous intéressons à la théorie spectrale et la théorie de la diffusion pour des systèmes d’opérateurs pseudodifférentiels auto-adjoints. Nous développons une approche stationnaire pour l’étude de la fonction de décalage spectral associée à une paire d’opérateurs de Schrödinger semi-classiques à potentiels matriciels. Une asymptotique de type Weyl avec reste optimal sur la fonction de décalage spectral est établie, et sous l’hypothèse d’existence d’une fonction fuite scalaire, un développement asymptotique complet en puissancesde h au sens fort sur sa dérivée est obtenu. Ce dernier résultat est une généralisation au cas matriciel d’un résultat de Robert et Tamura établi dans le cas scalaire près des énergies non-captives. Notre méthode indépendante du temps nous permet de traiter certains potentiels avec des croisements des valeurs propres. / In this work, we are interested in the spectral analysis of systems of semiclassical pseudodifferentialoperators. In the first part, we study the extension of the long time semiclassical Egorovtheorem in the case where the quantum Hamiltonian which generates the time evolution andthe initial quantum observable are two semiclassical pseudodifferential operators with matrixvaluedsymbols. Under an hyperbolicity condition on the principal symbol of the Hamiltonianwhich ensures the existence of the semiclassical projections, and for a class of observable thatare "semi-classically" block-diagonal with respect to these projections, we prove an Egorov theoremvalid in a large time interval of order log(h-1) known as the Ehrenfest time. Here h & 0is the semiclassical parameter.In the second part, we are interested in the spectral and scattering theories for self-adjointsystems of pseudodifferential operators. We develop a stationary approach for the study of thespectral shift function (SSF) associated to a pair of self-adjoint semiclassical Schrödinger operatorswith matrix-valued potentials. We prove a Weyl-type asymptotics with sharp remainderestimate on the SSF, and under the existence of a scalar escape function, a pointwise completeasymptotic expansion on its derivative. This last result is a generalisation in the matrix-valuedcase of a result of Robert and Tamura established in the scalar case near non-trapping energies.Our time-independent method allows us to treat certain potentials with energy-level crossings

La géométrie complexe est un outil puissant pour étudier les systèmes intégrables classiques, la physique statistique sur réseau aléatoire, les problèmes de matrices aléatoires, la théorie topologique des cordes, …Tous ces problèmes ont en commun la présence de relations, appelées équations de boucle ou contraintes de Virasoro. Dans le cas le plus simple, leur solution complète a été trouvée récemment, et se formule naturellement en termes de géométrie différentielle sur une surface de Riemann : la "courbe spectrale", qui dépend du problème. Cette thèse est une contribution au développement de ces techniques et de leurs applications.Pour commencer, nous abordons les questions de développement asymptotique à tous les ordres lorsque N tend vers l’infini, des intégrales N-dimensionnelles venant de la théorie des matrices aléatoires de taille N par N, ou plus généralement des gaz de Coulomb. Nous expliquons comment établir, dans les modèles de matrice beta et dans un régime à une coupure, le développement asymptotique à tous les ordres en puissances de N. Nous appliquons ces résultats à l'étude des grandes déviations du maximum des valeurs propres dans les modèles beta, et en déduisons de façon heuristique des informations sur l'asymptotique à tous les ordres de la loi de Tracy-Widom beta, pour tout beta positif. Ensuite, nous examinons le lien entre intégrabilité et équations de boucle. En corolaire, nous pouvons démontrer l'heuristique précédente concernant l'asymptotique de la loi de Tracy-Widom pour les matrices hermitiennes.Nous terminons avec la résolution de problèmes combinatoires en toute topologie. En théorie topologique des cordes, une conjecture de Bouchard, Klemm, Mariño et Pasquetti affirme que des séries génératrices bien choisies d'invariants de Gromov-Witten dans les espaces de Calabi-Yau toriques, sont solution d'équations de boucle. Nous l'avons démontré dans le cas le plus simple, où ces invariants coïncident avec les nombres de Hurwitz simples. Nous expliquons les progrès récents vers la conjecture générale, en relation avec nos travaux. En physique statistique sur réseau aléatoire, nous avons résolu le modèle O(n) trivalent sur réseau aléatoire introduit par Kostov, et expliquons la démarche à suivre pour résoudre des modèles plus généraux.Tous ces travaux soulignent l'importance de certaines "intégrales de matrices généralisées" pour les applications futures. Nous indiquons quelques éléments appelant à une théorie générale, encore basée sur des "équations de boucles", pour les calculer / Complex analysis is a powerful tool to study classical integrable systems, statistical physics on the random lattice, random matrix theory, topological string theory, … All these topics share certain relations, called "loop equations" or "Virasoro constraints". In the simplest case, the complete solution of those equations was found recently : it can be expressed in the framework of differential geometry over a certain Riemann surface which depends on the problem : the "spectral curve". This thesis is a contribution to the development of these techniques, and to their applications.First, we consider all order large N asymptotics in some N-dimensional integrals coming from random matrix theory, or more generally from "log gases" problems. We shall explain how to use loop equations to establish those asymptotics in beta matrix models within a one cut regime. This can be applied in the study of large fluctuations of the maximum eigenvalue in beta matrix models, and lead us to heuristic predictions about the asymptotics of Tracy-Widom beta law to all order, and for all positive beta. Second, we study the interplay between integrability and loop equations. As a corollary, we are able to prove the previous prediction about the asymptotics to all order of Tracy-Widom law for hermitian matrices.We move on with the solution of some combinatorial problems in all topologies. In topological string theory, a conjecture from Bouchard, Klemm, Mariño and Pasquetti states that certain generating series of Gromov-Witten invariants in toric Calabi-Yau threefolds, are solutions of loop equations. We have proved this conjecture in the simplest case, where those invariants coincide with the "simple Hurwitz numbers". We also explain recent progress towards the general conjecture, in relation with our work. In statistical physics on the random lattice, we have solved the trivalent O(n) model introduced by Kostov, and we explain the method to solve more general statistical models.Throughout the thesis, the computation of some "generalized matrices integrals" appears to be increasingly important for future applications, and this appeals for a general theory of loop equations.

No description available.

Submitted by Paulo Gustavo Grahl (pgrahl@fgvmail.br) on 2013-10-18T05:32:44Z No. of bitstreams: 1 DoutoradoPG_final.pdf: 23501670 bytes, checksum: 55b15051b9acc69ac74e639efe776fae (MD5) / Approved for entry into archive by ÁUREA CORRÊA DA FONSECA CORRÊA DA FONSECA (aurea.fonseca@fgv.br) on 2013-10-28T18:22:53Z (GMT) No. of bitstreams: 1 DoutoradoPG_final.pdf: 23501670 bytes, checksum: 55b15051b9acc69ac74e639efe776fae (MD5) / Approved for entry into archive by Marcia Bacha (marcia.bacha@fgv.br) on 2013-10-29T18:24:15Z (GMT) No. of bitstreams: 1 DoutoradoPG_final.pdf: 23501670 bytes, checksum: 55b15051b9acc69ac74e639efe776fae (MD5) / Made available in DSpace on 2013-10-29T18:25:35Z (GMT). No. of bitstreams: 1 DoutoradoPG_final.pdf: 23501670 bytes, checksum: 55b15051b9acc69ac74e639efe776fae (MD5) Previous issue date: 2012-12-22 / Esta dissertação concentra-se nos processos estocásticos espaciais definidos em um reticulado, os chamados modelos do tipo Cliff & Ord. Minha contribuição nesta tese consiste em utilizar aproximações de Edgeworth e saddlepoint para investigar as propriedades em amostras finitas do teste para detectar a presença de dependência espacial em modelos SAR (autoregressivo espacial), e propor uma nova classe de modelos econométricos espaciais na qual os parâmetros que afetam a estrutura da média são distintos dos parâmetros presentes na estrutura da variância do processo. Isto permite uma interpretação mais clara dos parâmetros do modelo, além de generalizar uma proposta de taxonomia feita por Anselin (2003). Eu proponho um estimador para os parâmetros do modelo e derivo a distribuição assintótica do estimador. O modelo sugerido na dissertação fornece uma interpretação interessante ao modelo SARAR, bastante comum na literatura. A investigação das propriedades em amostras finitas dos testes expande com relação a literatura permitindo que a matriz de vizinhança do processo espacial seja uma função não-linear do parâmetro de dependência espacial. A utilização de aproximações ao invés de simulações (mais comum na literatura), permite uma maneira fácil de comparar as propriedades dos testes com diferentes matrizes de vizinhança e corrigir o tamanho ao comparar a potência dos testes. Eu obtenho teste invariante ótimo que é também localmente uniformemente mais potente (LUMPI). Construo o envelope de potência para o teste LUMPI e mostro que ele é virtualmente UMP, pois a potência do teste está muito próxima ao envelope (considerando as estruturas espaciais definidas na dissertação). Eu sugiro um procedimento prático para construir um teste que tem boa potência em uma gama de situações onde talvez o teste LUMPI não tenha boas propriedades. Eu concluo que a potência do teste aumenta com o tamanho da amostra e com o parâmetro de dependência espacial (o que está de acordo com a literatura). Entretanto, disputo a visão consensual que a potência do teste diminui a medida que a matriz de vizinhança fica mais densa. Isto reflete um erro de medida comum na literatura, pois a distância estatística entre a hipótese nula e a alternativa varia muito com a estrutura da matriz. Fazendo a correção, concluo que a potência do teste aumenta com a distância da alternativa à nula, como esperado. / This dissertation focus on spatial stochastic process on a lattice (Cliff & Ord--type of models). My contribution consists of using Edgeworth and saddlepoint series to investigate small sample size and power properties of tests for detecting spatial dependence in spatial autoregressive (SAR) stochastic processes, and proposing a new class of spatial econometric models where the spatial dependence parameters that enter the mean structure are different from the ones in the covariance structure. This allows a clearer interpretation of models' parameters and generalizes the set of local and global models suggested by Anselin (2003) as an alternative to the traditional Cliff & Ord models. I propose an estimation procedure for the model's parameters and derive the asymptotic distribution of the parameters' estimators. The suggested model provides some insights on the structure of the commonly used mixed regressive, spatial autoregressive model with spatial autoregressive disturbances (SARAR). The study of the small sample properties of tests to detect spatial dependence expands on the existing literature by allowing the neighborhood structure to be a nonlinear function of the spatial dependence parameter. The use of series approximations instead of the often used Monte Carlo simulation allows a simple way to compare test properties across different neighborhood structures and to correct for size when comparing power. I obtain the power envelope for testing the presence of spatial dependence in the SAR process using the optimal invariant test statistic, which is also locally uniformly most powerful invariant (LUMPI). I have found that the LUMPI test is virtually UMP since its power is very close to the power envelope. I suggest a practical procedure to build a test that, while not UMP, retain good power properties in a wider range for the spatial parameter when compared to the LUMPI test. I find that power increases with sample size and with the spatial dependence parameter -- which agrees with the literature. However, I call into question the consensus view that power decreases as the spatial weight matrix becomes more densely connected. This finding in the literature reflects an error of measure because the hypothesis being compared are at very different statistical distance from the null. After adjusting for this, the power is larger for alternative hypothesis further away from the null -- as one would expect.