This thesis proposes new analysis tools for simulation models in the presence of data. To achieve a representation close to reality, simulation models are typically endowed with a set of inputs, termed parameters, that represent several controllable, stochastic or unknown components of the system. Because these models often utilize computationally expensive procedures, even modern supercomputers require a nontrivial amount of time, money, and energy to run for complex systems. Existing statistical frameworks avoid repeated evaluations of deterministic models through an emulator, constructed by conducting an experiment on the code. In high dimensional scenarios, the traditional framework for emulator-based analysis can fail due to the computational burden of inference. This thesis proposes a new class of experiments where inference from half a million observations is possible in seconds versus the days required for the traditional technique. In a case study presented in this thesis, the parameter of interest is a function as opposed to a scalar or a set of scalars, meaning the problem exists in the high dimensional regime. This work develops a new modeling strategy to nonparametrically study the functional parameter using Bayesian inference. Stochastic simulations are also investigated in the thesis. I describe the development of emulators through a framework termed quantile kriging, which allows for non-parametric representations of the stochastic behavior of the output whereas previous work has focused on normally distributed outputs. Furthermore, this work studied asymptotic properties of this methodology that yielded practical insights. Under certain regulatory conditions, there is the following result: By using an experiment that has the appropriate ratio of replications to sets of different inputs, we can achieve an optimal rate of convergence. Additionally, this method provided the basic tool for the study of defect patterns and a case study is explored.

Let $\Omega\subset\mathbb{R}^n$ be an open and connected subset of $\mathbb{R}^n$. We say that a function $F\colon \Omega-\Omega\to\mathbb{C}$, where $\Omega-\Omega=\{x-y\colon x,y\in\Omega\}$, is positive definite if for any $x_1,\ldots,x_m\in\Omega$ and any $c_1,\ldots,c_m\in \mathbb{C}$ we have that $\sum_{j,k=1}^m F(x_j-x_k)c_j\overline{c_k}\geq 0$. Let $F\colon\Omega-\Omega\to\mathbb{C}$ be a continuous positive definite function. We give necessary and sufficient conditions for $F$ to have an extension to a continuous and positive definite function defined on the entire Euclidean space $\mathbb{R}^n$. The conditions are formulated in terms of strong commutativity of some certain selfadjoint operators defined on a Hilbert space associated to our positive definite function.

In the first part of this paper we show a similarity between the principle of Structural Risk Minimization Principle (SRM) (Vapnik, 1982) and the idea of Sparse Approximation, as defined in (Chen, Donoho and Saunders, 1995) and Olshausen and Field (1996). Then we focus on two specific (approximate) implementations of SRM and Sparse Approximation, which have been used to solve the problem of function approximation. For SRM we consider the Support Vector Machine technique proposed by V. Vapnik and his team at AT&T Bell Labs, and for Sparse Approximation we consider a modification of the Basis Pursuit De-Noising algorithm proposed by Chen, Donoho and Saunders (1995). We show that, under certain conditions, these two techniques are equivalent: they give the same solution and they require the solution of the same quadratic programming problem.

No description available.

This thesis aims to contribute to the area of kernel methods, which are a class of machine learning methods known for their wide applicability and state-of-the-art performance, but which suffer from high training and evaluation complexity. The work in this thesis utilizes the notion of reduced-set models to alleviate the training and testing complexities of these methods in a unified manner. In the first part of the thesis, we use recent results in kernel smoothing and integral-operator learning to design a generic strategy to speed up various kernel methods. In Chapter 3, we present a method to speed up kernel PCA (KPCA), which is one of the fundamental kernel methods for manifold learning, by using reduced-set density estimates (RSDE) of the data. The proposed method induces an integral operator that is an approximation of the ideal integral operator associated to KPCA. It is shown that the error between the ideal and approximate integral operators is related to the error between the ideal and approximate kernel density estimates of the data. In Chapter 4, we derive similar approximation algorithms for Gaussian process regression, diffusion maps, and kernel embeddings of conditional distributions. In the second part of the thesis, we use reduced-set models for kernel methods to tackle online learning in model-reference adaptive control (MRAC). In Chapter 5, we relate the properties of the feature spaces induced by Mercer kernels to make a connection between persistency-of-excitation and the budgeted placement of kernels to minimize tracking and modeling error. In Chapter 6, we use a Gaussian process (GP) formulation of the modeling error to accommodate a larger class of errors, and design a reduced-set algorithm to learn a GP model of the modeling error. Proofs of stability for all the algorithms are presented, and simulation results on a challenging control problem validate the methods.

Este trabalho é dedicado ao estudo de propriedades teóricas dos operadores integrais positivos em \'L POT. 2\' (X; u), quando X é um espaço topológico localmente compacto ou primeiro enumerável e u é uma medida estritamente positiva. Damos ênfase à análise de propriedades espectrais relacionadas com extensões do Teorema de Mercer e ao estudo dos espaços de Hilbert de reprodução relacionados. Como aplicação, estudamos o decaimento dos autovalores destes operadores, em um contexto especial. Finalizamos o trabalho com a análise de propriedades de suavidade das funções do espaço de Hilbert de reprodução, quando X é um subconjunto do espaço euclidiano usual e u é a medida de Lebesgue usual de X / In this work we study theoretical properties of positive integral operators on \'L POT. 2\'(X; u), in the case when X is a topological space, either locally compact or first countable, and u is a strictly positive measure. The analysis is directed to spectral properties of the operator which are related to some extensions of Mercer\'s Theorem and to the study of the reproducing kernel Hilbert spaces involved. As applications, we deduce decay rates for the eigenvalues of the operators in a special but relevant case. We also consider smoothness properties for functions in the reproducing kernel Hilbert spaces when X is a subset of the Euclidean space and u is the Lebesgue measure of the space

Neste trabalho analisamos o papel das funções layout de um núcleo positivo definido K sobre um espaço topológico de Hausdor E com relação a duas propriedades específicas: a universalidade de K e a ortogonalidade no espaço de Hilbert de reprodução de K a partir de suportes disjuntos. As funções layout sempre existem mas podem não ser únicas. De uma maneira geral, a função layout e uma aplicação que transfere, convenientemente, informações do espaço E para um espaço com produto interno de dimensão alta, onde métodos lineares podem ser usados. Tanto a universalidade quanto a ortogonalidade pressupõem a continuidade do núcleo. O primeiro conceito exige que para cada compacto não vazio X de E, o conjunto de \"seções\" {K(., y) : y \'PERTENCE\' X} seja total no espaço de todas as funções contínuas com domínio X, munido da topologia da convergência uniforme. Um dos resultados principais do trabalho caracteriza a universalidade de um núcleo K através de uma propriedade de universalidade semelhante da função layout. A ortogonalidade a partir de suportes disjuntos almeja então a ortogonalidade de quaisquer duas funções do espaço de Hilbert de reprodução de K quando seus suportes não se intersectam / We analyze the role of feature maps of a positive denite kernel K acting on a Hausdorff topological space E in two specific properties: the universality of K and the orthogonality in the reproducing kernel Hilbert space of K from disjoint supports. Feature maps always exist but may not be unique. A feature map may be interpreted as a kernel based procedure that maps the data from the original input space E into a potentially higher dimensional \"feature space\" in which linear methods may then be used. Both properties, universality and orthogonality from disjoint supports, make sense under continuity of the kernel. Universality of K is equivalent to the fundamentality of {K(. ; y) : y \'IT BELONGS\' X} in the space of all continuous functions on X, with the topology of uniform convergence, for all nonempty compact subsets X of E. One of the main results in this work is a characterization of the universality of K from a similar concept for the feature map. Orthogonality from disjoint supports seeks the orthogonality of any two functions in the reproducing kernel Hilbert space of K when the functions have disjoint supports

Este trabalho é dedicado ao estudo de propriedades teóricas dos operadores integrais positivos em \'L POT. 2\' (X; u), quando X é um espaço topológico localmente compacto ou primeiro enumerável e u é uma medida estritamente positiva. Damos ênfase à análise de propriedades espectrais relacionadas com extensões do Teorema de Mercer e ao estudo dos espaços de Hilbert de reprodução relacionados. Como aplicação, estudamos o decaimento dos autovalores destes operadores, em um contexto especial. Finalizamos o trabalho com a análise de propriedades de suavidade das funções do espaço de Hilbert de reprodução, quando X é um subconjunto do espaço euclidiano usual e u é a medida de Lebesgue usual de X / In this work we study theoretical properties of positive integral operators on \'L POT. 2\'(X; u), in the case when X is a topological space, either locally compact or first countable, and u is a strictly positive measure. The analysis is directed to spectral properties of the operator which are related to some extensions of Mercer\'s Theorem and to the study of the reproducing kernel Hilbert spaces involved. As applications, we deduce decay rates for the eigenvalues of the operators in a special but relevant case. We also consider smoothness properties for functions in the reproducing kernel Hilbert spaces when X is a subset of the Euclidean space and u is the Lebesgue measure of the space

Neste trabalho analisamos o papel das funções layout de um núcleo positivo definido K sobre um espaço topológico de Hausdor E com relação a duas propriedades específicas: a universalidade de K e a ortogonalidade no espaço de Hilbert de reprodução de K a partir de suportes disjuntos. As funções layout sempre existem mas podem não ser únicas. De uma maneira geral, a função layout e uma aplicação que transfere, convenientemente, informações do espaço E para um espaço com produto interno de dimensão alta, onde métodos lineares podem ser usados. Tanto a universalidade quanto a ortogonalidade pressupõem a continuidade do núcleo. O primeiro conceito exige que para cada compacto não vazio X de E, o conjunto de \"seções\" {K(., y) : y \'PERTENCE\' X} seja total no espaço de todas as funções contínuas com domínio X, munido da topologia da convergência uniforme. Um dos resultados principais do trabalho caracteriza a universalidade de um núcleo K através de uma propriedade de universalidade semelhante da função layout. A ortogonalidade a partir de suportes disjuntos almeja então a ortogonalidade de quaisquer duas funções do espaço de Hilbert de reprodução de K quando seus suportes não se intersectam / We analyze the role of feature maps of a positive denite kernel K acting on a Hausdorff topological space E in two specific properties: the universality of K and the orthogonality in the reproducing kernel Hilbert space of K from disjoint supports. Feature maps always exist but may not be unique. A feature map may be interpreted as a kernel based procedure that maps the data from the original input space E into a potentially higher dimensional \"feature space\" in which linear methods may then be used. Both properties, universality and orthogonality from disjoint supports, make sense under continuity of the kernel. Universality of K is equivalent to the fundamentality of {K(. ; y) : y \'IT BELONGS\' X} in the space of all continuous functions on X, with the topology of uniform convergence, for all nonempty compact subsets X of E. One of the main results in this work is a characterization of the universality of K from a similar concept for the feature map. Orthogonality from disjoint supports seeks the orthogonality of any two functions in the reproducing kernel Hilbert space of K when the functions have disjoint supports

Le but de l’apprentissage supervisé est d’inférer des relations entre un phénomène que l’on souhaite prédire et des variables « explicatives ». À cette fin, on dispose d’observations de multiples réalisations du phénomène, à partir desquelles on propose une règle de prédiction. L’émergence récente de sources de données à très grande échelle, tant par le nombre d’observations effectuées (en analyse d’image, par exemple) que par le grand nombre de variables explicatives (en génétique), a fait émerger deux difficultés : d’une part, il devient difficile d’éviter l’écueil du sur-apprentissage lorsque le nombre de variables explicatives est très supérieur au nombre d’observations; d’autre part, l’aspect algorithmique devient déterminant, car la seule résolution d’un système linéaire dans les espaces en jeupeut devenir une difficulté majeure. Des algorithmes issus des méthodes d’approximation stochastique proposent uneréponse simultanée à ces deux difficultés : l’utilisation d’une méthode stochastique réduit drastiquement le coût algorithmique, sans dégrader la qualité de la règle de prédiction proposée, en évitant naturellement le sur-apprentissage. En particulier, le cœur de cette thèse portera sur les méthodes de gradient stochastique. Les très populaires méthodes paramétriques proposent comme prédictions des fonctions linéaires d’un ensemble choisi de variables explicatives. Cependant, ces méthodes aboutissent souvent à une approximation imprécise de la structure statistique sous-jacente. Dans le cadre non-paramétrique, qui est un des thèmes centraux de cette thèse, la restriction aux prédicteurs linéaires est levée. La classe de fonctions dans laquelle le prédicteur est construit dépend elle-même des observations. En pratique, les méthodes non-paramétriques sont cruciales pour diverses applications, en particulier pour l’analyse de données non vectorielles, qui peuvent être associées à un vecteur dans un espace fonctionnel via l’utilisation d’un noyau défini positif. Cela autorise l’utilisation d’algorithmes associés à des données vectorielles, mais exige une compréhension de ces algorithmes dans l’espace non-paramétrique associé : l’espace à noyau reproduisant. Par ailleurs, l’analyse de l’estimation non-paramétrique fournit également un éclairage révélateur sur le cadre paramétrique, lorsque le nombre de prédicteurs surpasse largement le nombre d’observations. La première contribution de cette thèse consiste en une analyse détaillée de l’approximation stochastique dans le cadre non-paramétrique, en particulier dans le cadre des espaces à noyaux reproduisants. Cette analyse permet d’obtenir des taux de convergence optimaux pour l’algorithme de descente de gradient stochastique moyennée. L’analyse proposée s’applique à de nombreux cadres, et une attention particulière est portée à l’utilisation d’hypothèses minimales, ainsi qu’à l’étude des cadres où le nombre d’observations est connu à l’avance, ou peut évoluer. La seconde contribution est de proposer un algorithme, basé sur un principe d’accélération, qui converge à une vitesse optimale, tant du point de vue de l’optimisation que du point de vue statistique. Cela permet, dans le cadre non-paramétrique, d’améliorer la convergence jusqu’au taux optimal, dans certains régimes pour lesquels le premier algorithme analysé restait sous-optimal. Enfin, la troisième contribution de la thèse consiste en l’extension du cadre étudié au delà de la perte des moindres carrés : l’algorithme de descente de gradient stochastiqueest analysé comme une chaine de Markov. Cette approche résulte en une interprétation intuitive, et souligne les différences entre le cadre quadratique et le cadre général. Une méthode simple permettant d’améliorer substantiellement la convergence est également proposée. / The goal of supervised machine learning is to infer relationships between a phenomenon one seeks to predict and “explanatory” variables. To that end, multiple occurrences of the phenomenon are observed, from which a prediction rule is constructed. The last two decades have witnessed the apparition of very large data-sets, both in terms of the number of observations (e.g., in image analysis) and in terms of the number of explanatory variables (e.g., in genetics). This has raised two challenges: first, avoiding the pitfall of over-fitting, especially when the number of explanatory variables is much higher than the number of observations; and second, dealing with the computational constraints, such as when the mere resolution of a linear system becomes a difficulty of its own. Algorithms that take their roots in stochastic approximation methods tackle both of these difficulties simultaneously: these stochastic methods dramatically reduce the computational cost, without degrading the quality of the proposed prediction rule, and they can naturally avoid over-fitting. As a consequence, the core of this thesis will be the study of stochastic gradient methods. The popular parametric methods give predictors which are linear functions of a set ofexplanatory variables. However, they often result in an imprecise approximation of the underlying statistical structure. In the non-parametric setting, which is paramount in this thesis, this restriction is lifted. The class of functions from which the predictor is proposed depends on the observations. In practice, these methods have multiple purposes, and are essential for learning with non-vectorial data, which can be mapped onto a vector in a functional space using a positive definite kernel. This allows to use algorithms designed for vectorial data, but requires the analysis to be made in the non-parametric associated space: the reproducing kernel Hilbert space. Moreover, the analysis of non-parametric regression also sheds some light on the parametric setting when the number of predictors is much larger than the number of observations. The first contribution of this thesis is to provide a detailed analysis of stochastic approximation in the non-parametric setting, precisely in reproducing kernel Hilbert spaces. This analysis proves optimal convergence rates for the averaged stochastic gradient descent algorithm. As we take special care in using minimal assumptions, it applies to numerous situations, and covers both the settings in which the number of observations is known a priori, and situations in which the learning algorithm works in an on-line fashion. The second contribution is an algorithm based on acceleration, which converges at optimal speed, both from the optimization point of view and from the statistical one. In the non-parametric setting, this can improve the convergence rate up to optimality, even inparticular regimes for which the first algorithm remains sub-optimal. Finally, the third contribution of the thesis consists in an extension of the framework beyond the least-square loss. The stochastic gradient descent algorithm is analyzed as a Markov chain. This point of view leads to an intuitive and insightful interpretation, that outlines the differences between the quadratic setting and the more general setting. A simple method resulting in provable improvements in the convergence is then proposed.