<div>Uncertainty quantification techniques based on the spectral approach have been studied extensively in the literature to characterize and quantify, at low computational cost, the impact that uncertainties may have on large-scale engineering problems. One such technique is the <i>generalized polynomial chaos</i> (gPC) which utilizes a time-independent orthogonal basis to expand a stochastic process in the space of random functions. The method uses a specific Askey-chaos system that is concordant with the measure defined in the probability space in order to ensure exponential convergence to the solution. For nearly two decades, this technique has been used widely by several researchers in the area of uncertainty quantification to solve stochastic problems using the spectral approach. However, a major drawback of the gPC method is that it cannot be used in the resolution of problems that feature strong nonlinear dependencies over the probability space as time progresses. Such downside arises due to the time-independent nature of the random basis, which has the undesirable property to lose unavoidably its optimality as soon as the probability distribution of the system's state starts to evolve dynamically in time.</div><div><br></div><div>Another technique is the <i>time-dependent generalized polynomial chaos</i> (TD-gPC) which utilizes a time-dependent orthogonal basis to better represent the stochastic part of the solution space (aka random function space or RFS) in time. The development of this technique was motivated by the fact that the probability distribution of the solution changes with time, which in turn requires that the random basis is frequently updated during the simulation to ensure that the mean-square error is kept orthogonal to the discretized RFS. Though this technique works well for problems that feature strong nonlinear dependencies over the probability space, the TD-gPC method possesses a serious issue: it suffers from the curse of dimensionality at the RFS level. This is because in all gPC-based methods the RFS is constructed using a tensor product of vector spaces with each of these representing a single RFS over one of the dimensions of the probability space. As a result, the higher the dimensionality of the probability space, the more vector spaces needed in the construction of a suitable RFS. To reduce the dimensionality of the RFS (and thus, its associated computational cost), gPC-based methods require the use of versatile sparse tensor products within their numerical schemes to alleviate to some extent the curse of dimensionality at the RFS level. Therefore, this curse of dimensionality in the TD-gPC method alludes to the need of developing a more compelling spectral method that can quantify uncertainties in long-time response of dynamical systems at much lower computational cost.</div><div><br></div><div>In this work, a novel numerical method based on the spectral approach is proposed to resolve the curse-of-dimensionality issue mentioned above. The method has been called the <i>flow-driven spectral chaos</i> (FSC) because it uses a novel concept called <i>enriched stochastic flow maps</i> to track the evolution of a finite-dimensional RFS efficiently in time. The enriched stochastic flow map does not only push the system's state forward in time (as would a traditional stochastic flow map) but also its first few time derivatives. The push is performed this way to allow the random basis to be constructed using the system's enriched state as a germ during the simulation and so as to guarantee exponential convergence to the solution. It is worth noting that this exponential convergence is achieved in the FSC method by using only a few number of random basis vectors, even when the dimensionality of the probability space is considerably high. This is for two reasons: (1) the cardinality of the random basis does not depend upon the dimensionality of the probability space, and (2) the cardinality is bounded from above by <i>M+n+1</i>, where <i>M</i> is the order of the stochastic flow map and <i>n</i> is the order of the governing stochastic ODE. The boundedness of the random basis from above is what makes the FSC method be curse-of-dimensionality free at the RFS level. For instance, for a dynamical system that is governed by a second-order stochastic ODE (<i>n=2</i>) and driven by a stochastic flow map of fourth-order (<i>M=4</i>), the maximum number of random basis vectors to consider within the FSC scheme is just 7, independent whether the dimensionality of the probability space is as low as 1 or as high as 10,000.</div><div><br></div><div>With the aim of reducing the complexity of the presentation, this dissertation includes three levels of abstraction for the FSC method, namely: a <i>specialized version</i> of the FSC method for dealing with structural dynamical systems subjected to uncertainties (Chapter 2), a <i>generalized version</i> of the FSC method for dealing with dynamical systems governed by (nonlinear) stochastic ODEs of arbitrary order (Chapter 3), and a <i>multi-element version</i> of the FSC method for dealing with dynamical systems that exhibit discontinuities over the probability space (Chapter 4). This dissertation also includes an implementation of the FSC method to address the dynamics of large-scale stochastic structural systems more effectively (Chapter 5). The implementation is done via a modal decomposition of the spatial function space as a means to reduce the number of degrees of freedom in the system substantially, and thus, save computational runtime.</div>

Thesis (Ph. D.)--Joint Program in Applied Ocean Science and Engineering (Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science; and the Woods Hole Oceanographic Institution), 2006. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Also issued in pages. Barker Engineering Library copy: issued in pages. / Includes bibliographical references (leaves 193-[201]). / The first part of the dissertation investigates the application of the theory of large random matrices to high-dimensional inference problems when the samples are drawn from a multivariate normal distribution. A longstanding problem in sensor array processing is addressed by designing an estimator for the number of signals in white noise that dramatically outperforms that proposed by Wax and Kailath. This methodology is extended to develop new parametric techniques for testing and estimation. Unlike techniques found in the literature, these exhibit robustness to high-dimensionality, sample size constraints and eigenvector misspecification. By interpreting the eigenvalues of the sample covariance matrix as an interacting particle system, the existence of a phase transition phenomenon in the largest ("signal") eigenvalue is derived using heuristic arguments. This exposes a fundamental limit on the identifiability of low-level signals due to sample size constraints when using the sample eigenvalues alone. The analysis is extended to address a problem in sensor array processing, posed by Baggeroer and Cox, on the distribution of the outputs of the Capon-MVDR beamformer when the sample covariance matrix is diagonally loaded. / (cont.) The second part of the dissertation investigates the limiting distribution of the eigenvalues and eigenvectors of a broader class of random matrices. A powerful method is proposed that expands the reach of the theory beyond the special cases of matrices with Gaussian entries; this simultaneously establishes a framework for computational (non-commutative) "free probability" theory. The class of "algebraic" random matrices is defined and the generators of this class are specified. Algebraicity of a random matrix sequence is shown to act as a certificate of the computability of the limiting eigenvalue distribution and, for a subclass, the limiting conditional "eigenvector distribution." The limiting moments of algebraic random matrix sequences, when they exist, are shown to satisfy a finite depth linear recursion so that they may often be efficiently enumerated in closed form. The method is applied to predict the deterioration in the quality of the sample eigenvectors of large algebraic empirical covariance matrices due to sample size constraints. / by Rajesh Rao Nadakuditi. / Ph.D.

<p>In the real world, uncertainty is a common challenging problem faced by individuals, organizations, and firms. Decision quality is highly impacted by uncertainty because decision makers lack complete information and have to leverage the loss and gain in many possible outcomes or scenarios. This study explores dynamic decision making (with known distributions) and decision learning (with unknown distributions but some samples) in not-for-profit operations and supply chain management. We first study dynamic staffing for paid workers and volunteers with uncertain supply in a nonprofit operation where the optimal policy is too complex to compute and implement. Then, we consider dynamic inventory control and pricing under both supply and demand uncertainties where unmet demand is lost leading to a challenging non-concave dynamic problem. Furthermore, we explore decision learning from limited data of focal system and available data of related but different systems by transfer learning, cross learning, and co-learning utilizing the similarities among related systems.</p>

We present results that span three interconnected domains. Initially, our analysis is centred on Backward Stochastic Differential Equations (BSDEs) featuring time-delayed generators. Subsequently, we direct our interest towards Mean Field Games (MFGs) incorporating absorption aspects, with a focus on the corresponding Master Equation within a confined domain under the imposition of Dirichlet boundary conditions. The investigation culminates in exploring pertinent Machine Learning methodologies applied to financial and economic decision-making processes.

[ES] El tema de la investigación se centra en técnicas avanzadas para manejar problemas de aguas subterráneas y superficiales relacionados con métodos inversos y cambio climático. Los filtros de Kalman, con especial atención en Ensemble Smoother with Multiple Data Assimilation (ES-MDA), se analizan y mejoran para la solución de diferentes tipos de problemas inversos. En particular, la principal novedad es la aplicación de estos métodos para la identificación de series temporales. La primera parte de la tesis, luego de la descripción del método, presenta el desarrollo de un software escrito en Python para la aplicación de la metodología propuesta. El software cuenta con un flujo de trabajo flexible que puede adaptarse fácilmente para implementar diferentes variantes del filtro de Kalman y ser aplicado para la solución de varios tipos de problemas. Un paquete de herramientas proporciona varias funcionalidades que permiten de configurar el algoritmo de acuerdo con el problema específico analizado. La primera aplicación se refiere a la solución del problema inverso de flujo en ríos. Este es un procedimiento inverso destinado a estimar el flujo de entrada a un sistema hidráulico en función de información recopilada abajo. El procedimiento se prueba mediante dos ejemplos sintéticos y un estudio de caso real; se investiga el impacto de los tamaños de los conjuntos y la aplicación de técnicas de localización e inflación de covarianzas. Los resultados muestran la capacidad del método propuesto de resolver este tipo de problemas; el rendimiento de ES-MDA mejora, especialmente para tamaños de conjuntos pequeños, cuando se aplican técnicas de inflación y localización de covarianza. La segunda aplicación en el campo de las aguas superficiales se refiere a la calibración de un modelo hidrológico-hidráulico que simula los mecanismos de formación de eventos de inundación. ES-MDA se acopla al modelo numérico de forma paralela para la estimación de los coeficientes de rugosidad e infiltración en base al conocimiento de un hidrograma de flujo en una sección del dominio. Los resultados de dos casos sintéticos y un estudio de caso real demuestran la capacidad del método propuesto para calibrar el modelo hidrológico-hidráulico con un tiempo computacional razonable. En el campo de aguas subterráneas, ES-MDA se aplica por primera vez para identificar simultáneamente la ubicación de la fuente y el historial de liberación de un contaminante en un acuífero a partir de datos de concentración detectados en diferentes puntos del dominio. Se realizaron numerosas pruebas para evaluar la influencia de la distribución espacial y temporal de los datos de concentración, el número del conjunto y el uso de técnicas de localización e inflación; además, se presenta un nuevo procedimiento para realizar una localización iterativa espacio-temporal. La metodología se valida mediante un ejemplo analítico y un estudio de caso que utiliza datos obtenidos en el laboratorio mediante una caja de arena. ES-MDA conduce a una buena estimación de los parámetros investigados; una red de monitoreo bien diseñada y la aplicación de correcciones de covarianza mejoran el rendimiento del método y ayudan a mitigar el posible problema de no unicidad de la solución. Otro propósito de la tesis es investigar el efecto del cambio climático en las aguas subterráneas. Se presenta un modelo simplificado que describe la respuesta de los niveles de agua subterránea a las variables meteorológicas hasta 2100. Es un enfoque estadístico sencillo basado en las correlaciones entre los niveles de agua subterránea y dos índices de sequía que dependen de los datos de precipitación y temperatura. El método se utiliza para evaluar el impacto del cambio climático en los recursos de agua subterránea en un área de estudio ubicada en el norte de Italia utilizando datos históricos y de modelos climáticos regionales. Los resultados m / [CA] El tema de la investigació se centra en tècniques avançades per a manejar problemes d'aigües subterrànies i superficials relacionats amb mètodes inversos i canvi climàtic. Els filtres de Kalman, amb especial atenció en Ensemble Smoother with Multiple Data Assimilation (ES-MDA), s'analitzen i milloren per a la solució de diferents tipus de problemes inversos. En particular, la principal novetat és l'aplicació d'aquests mètodes per a la identificació de sèries temporals. La primera part de la tesi presenta el desenvolupament d'un programari escrit en Python per l'aplicació de la metodologia presentada. El programari compta amb un flux de treball flexible que pot adaptar-se fàcilment per a implementar diferents variants del filtre de Kalman i ser aplicat per a la solució de diversos tipus de problemes. Un paquet complementar d'eines proporciona diverses funcionalitats que permeten de configurar l'algorisme d'acord amb el problema específic analitzat. La primera aplicació es un nou enfocament per la solució del problema invers de flux en rius. Aquest és un procediment invers destinat a estimar el flux d'entrada a un sistema hidràulic en funció d'informació recopilada aigües avall. El procediment es prova mitjançant dos exemples sintètics i un estudi de cas real; s'investiga l'impacte de les grandàries dels conjunts i l'aplicació de tècniques de localització i inflació de covariàncies. Els resultats mostren la capacitat del mètode proposat de resoldre aquest tipus de problemes; el rendiment de ES-MDA millora, especialment per a grandàries de conjunts xicotets, quan s'apliquen tècniques d'inflació i localització de covariància. La segona aplicació en el camp de les aigües superficials es refereix al calibratge d'un model hidrològic-hidràulic que simula els mecanismes de formació d'esdeveniments d'inundació a partir de sollicitació hidrometeorológicas i la seua posterior propagació. ES-MDA s'acobla al model numèric de manera paral·lela per l'estimació dels coeficients de rugositat i infiltració sobre la base del coneixement d'un hidrograma de flux en una secció del domini. Els resultats de dos casos sintètics i un estudi de cas real demostren la capacitat del mètode proposat per calibrar el model hidrològic-hidràulic amb un temps computacional raonable. En el camp d'aigües subterrànies, ES-MDA s'aplica per primera vegada per identificar simultàniament la ubicació de la font i l'historial d'alliberament d'un contaminant en un aqüífer a partir d'un conjunt de dades de concentració detectats en diferents punts del domini. Es van realitzar nombroses proves per avaluar la influència de la distribució espacial i temporal de les dades de concentració, el número del conjunt i l'ús de tècniques de localització i inflació; a més, es presenta un nou procediment per realitzar una localització iterativa espaciotemporal. La metodologia es valguda mitjançant un exemple analític i un estudi de cas per al qual s'utilitzen dades obtingudes en el laboratori mitjançant una caixa d'arena. ES-MDA condueix a una bona reconstrucció dels paràmetres investigats; una xarxa de monitoratge ben dissenyada i l'aplicació de correccions de covariància milloren el rendiment del mètode i ajuden a mitigar el possible problema de no unicitat de la solució. Un altre propòsit de la tesi és investigar l'efecte del canvi climàtic en les aigües subterrànies. Es presenta un model simplificat que descriu la resposta dels nivells d'aigua subterrània a les variables meteorològiques fins a 2100. És un enfocament estadístic senzill basat en les correlacions entre els nivells d'aigua subterrània i dos índexs de sequera que depenen de les dades de precipitació i temperatura. El mètode s'utilitza per a avaluar l'impacte del canvi climàtic en els recursos d'aigua subterrània en una àrea d'estudi situada en el nord d'Itàlia utilitzant dades històriques i de models climàtics regionals. / [EN] This work focuses on the investigation of advanced techniques to handle groundwater and surface water problems in the framework of inverse methods and climate change. The Ensemble Kalman filter methods, with particular attention to the Ensemble Smoother with Multiple Data Assimilation (ES-MDA), are extensively analyzed and improved for the solution of different types of inverse problems. In particular, the main novelty is the application of these methods for the identification of time series function. In the first part of the thesis, after the description of the ES-MDA method, the development of a Python software package for the application of the proposed methodology is presented. It is designed with a flexible workflow that can be easily adapted to implement different variants of the Ensemble Kalman filter and to be applied for the solution of various types of inverse problems. A complemented tool package provides several functionalities that allow to setup the algorithm configuration suiting the specific analyzed problem. The first novelty application of the ES-MDA method aimed at solving the reverse flow routing problem. The objective of the inverse procedure is the estimation of an unknown inflow hydrograph to a hydraulic system on the basis of information collected downstream and a given forward routing model that relates inflow hydrograph and downstream observations. The procedure is tested by means of two synthetic examples and a real case study; the impact of ensemble sizes and the application of covariance localization and inflation techniques are also investigated. The tests show the capability of the proposed method to solve this type of problem; the performance of ES-MDA improves, especially for small ensemble sizes, when covariance localization and inflation techniques are applied. The second application, in the context of surface water, concerns the calibration of a hydrological-hydraulic model that simulates rainfall-runoff processes. The ES-MDA is coupled with the numerical model by parallel way for the estimation of roughness and infiltration coefficients based on the knowledge of a discharge hydrograph at the basin outlet. The results of two synthetic tests and a real case study demonstrate the capability of the proposed method to calibrate the hydrological-hydraulic model with a reasonable computational time. In the groundwater field, ES-MDA is applied for the first time to simultaneously identify the source location and the release history of a contaminant spill in an aquifer from a sparse set of concentration data collected in few points of the aquifer. The impacts of the concentration sampling scheme, the ensemble size and the use of covariance localization and covariance inflation techniques are tested; furthermore, a new procedure to perform a spatiotemporal iterative localization is presented. The methodology is tested by means of an analytical example and a study case that uses real data collected in a laboratory sandbox. ES-MDA leads to a good estimation of the investigated parameters; a well-designed monitoring network and the use of covariance corrections improve the performance of the method and help to minimize ill-posedness and equifinality. A part of the thesis investigates the impact of climate change on the groundwater availability. A surrogate model that describes the response of groundwater levels to meteorological variables up to 2100 is presented. It is a simple statistical approach based on the correlations between groundwater levels and two drought indices that depend on precipitation and temperature data. The presented method is used to evaluate the impact of climate change on groundwater resources in a study area located in Northern Italy using historical and regional climate model data. The results denote a progressive increase of groundwater droughts in the investigated area. / Todaro, V. (2021). Advanced techniques for solving groundwater and surface water problems in the context of inverse methods and climate change [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/166439

The overarching goal of this thesis is to use distance covariance based methods to extend asymptotic results from the i.i.d. case to general time series settings. Accounting for dependence may make already difficult statistical inference all the more challenging. The distance covariance is an increasingly popular measure of dependence between random vectors that goes beyond linear dependence as described by correlation. It is defined by a squared integral norm of the difference between the joint and marginal characteristic functions with respect to a specific weight function. Distance covariance has the advantage of being able to detect dependence even for uncorrelated data. The energy distance is a closely related quantity that measures distance between distributions of random vectors. These statistics can be used to establish asymptotic limit theory for stationary ergodic time series. The asymptotic results are driven by the limit theory for the empirical characteristic functions. In this thesis we apply the distance covariance to three problems in time series modeling: (i) Independent Component Analysis (ICA), (ii) multivariate time series clustering, and (iii) goodness-of-fit using residuals from a fitted model. The underlying statistical procedures for each topic uses the distance covariance function as a measure of dependence. The distance covariance arises in various ways in each of these topics; one as a measure of independence among the components of a vector, second as a measure of similarity of joint distributions and, third for assessing serial dependence among the fitted residuals. In each of these cases, limit theory is established for the corresponding empirical distance covariance statistics when the data comes from a stationary ergodic time series. For Topic (i) we consider an ICA framework, which is a popular tool used for blind source separation and has found application in fields such as financial time series, signal processing, feature extraction, and brain imaging. The Structural Vector Autogregression (SVAR) model is often the basic model used for modeling macro time series. The residuals in such a model are given by e_t = A S_t, the classical ICA model. In certain applications, one of the components of S_t has infinite variance. This differs from the standard ICA model. Furthermore the e_t's are not observed directly but are only estimated from the SVAR modeling. Many of the ICA procedures require the existence of a finite second or even fourth moment. We derive consistency when using the distance covariance for measuring independence of residuals under the infinite variance case.Extensions to the ICA model with noise, which has a direct application to SVAR models when testing independence of residuals based on their estimated counterparts is also considered. In Topic (ii) we propose a novel methodology for clustering multivariate time series data using energy distance. Specifically, a dissimilarity matrix is formed using the energy distance statistic to measure separation between the finite dimensional distributions for the component time series. Once the pairwise dissimilarity matrix is calculated, a hierarchical clustering method is then applied to obtain the dendrogram. This procedure is completely nonparametric as the dissimilarities between stationary distributions are directly calculated without making any model assumptions. In order to justify this procedure, asymptotic properties of the energy distance estimates are derived for general stationary and ergodic time series. Topic (iii) considers the fundamental and often final step in time series modeling, assessing the quality of fit of a proposed model to the data. Since the underlying distribution of the innovations that generate a model is often not prescribed, goodness-of-fit tests typically take the form of testing the fitted residuals for serial independence. However, these fitted residuals are inherently dependent since they are based on the same parameter estimates and thus standard tests of serial independence, such as those based on the autocorrelation function (ACF) or distance correlation function (ADCF) of the fitted residuals need to be adjusted. We apply sample splitting in the time series setting to perform tests of serial dependence of fitted residuals using the sample ACF and ADCF. Here the first f_n of the n data points in the time series are used to estimate the parameters of the model. Tests for serial independence are then based on all the n residuals. With f_n = n/2 the ACF and ADCF tests of serial independence tests often have the same limit distributions as though the underlying residuals are indeed i.i.d. That is, if the first half of the data is used to estimate the parameters and the estimated residuals are computed for the entire data set based on these parameter estimates, then the ACF and ADCF can have the same limit distributions as though the residuals were i.i.d. This procedure ameliorates the need for adjustment in the construction of confidence bounds for both the ACF and ADCF, based on the fitted residuals, in goodness-of-fit testing. We also show that if f_n < n/2 then the asymptotic distribution of the tests stochastically dominate the corresponding asymptotic distributions for the true i.i.d. noise; the stochastic order gets reversed under f_n > n/2.

Cette thèse traite de deux sujets: la résolubilité forte d'équations différentielles stochastiques à dérive hölderienne et bruit hypoelliptique et la simulation de processus progressifs-rétrogrades découplés de McKean-Vlasov. Dans le premier cas, on montre qu'un système hypoelliptique, composé d'une composante diffusive et d'une composante totalement dégénérée, est fortement résoluble lorsque l'exposant de la régularité Hölder de la dérive par rapport à la composante dégénérée est strictement supérieur à 2/3. Ce travail étend au cadre dégénéré les travaux antérieurs de Zvonkin (1974), Veretennikov (1980) et Krylov et Röckner (2005). L'apparition d'un seuil critique pour l'exposant peut-être vue comme le prix à payer pour la dégénérescence. La preuve repose sur des résultats de régularité de la solution de l'EDP associée, qui est dégénérée, et est basée sur une méthode parametrix. Dans le second cas, on propose un algorithme basé sur les méthodes de cubature pour la simulation de processus progessifs-rétrogrades découplés de McKean-Vlasov apparaissant dans des problèmes de contrôle dans un environnement de type champ moyen. Cet algorithme se divise en deux parties. Une première étape de construction d'un arbre de particules, à dynamique déterministe, approchant la loi de la composante progressive. Cet arbre peut être paramétré de manière à obtenir n'importe quel ordre d'approximation (en terme de pas de discrétisation de l'intervalle). Une seconde étape, conditionnelle à l'arbre, permettant l'approximation de la composante rétrograde. Deux schémas explicites sont proposés permettant un ordre d'approximation de 1 et 2. / This thesis deals with two subjects: the strong well-posedness of stochastic differential equations with Hölder drift and hypoelliptic noise and the simulation of decoupled forward backward stochastic differential equations of McKean-Vlasov type. In the first work, we study a class of degenerate system with hypoelliptic noise. We prove that strong well-posedness holds for this system when the drift is only H\"{o}lder, with Hölder exponent larger than the critical value 2/3. This work extends to the degenerate setting the earlier results obtained by Zvonkin (1974), Veretennikov (1980) and Krylov and Röckner (2005). The existence of a threshold for the Hölder exponent in the degenerate case may be understood as the price to pay to balance the degeneracy of the noise. Our proof relies on regularization properties of the associated PDE, which is degenerate in the current framework and is based on a parametrix method. In the second work, we propose a new algorithm to approach weakly the solution of a McKean-Vlasov stochastic differential equation. Based on the cubature method, the algorithm is deterministic differing from the usual methods based on interacting particles. It can be parametrized in order to obtain a given order of convergence. Then, we construct implementable algorithms to solve decoupled forward backward stochastic differential equations of McKean-Vlasov type, which appear in some stochastic control problems in a mean field environment. We give two algorithms and show that they have convergence of orders one and two under appropriate regularity conditions.

Diese Dissertation behandelt Fragen aus der stochastischen Analysis und der Finanzmathematik, die sich unter dem Begriff der Robustheit zusammenfassen lassen. Zunächst betrachten wir finanzmathematische Modelle mit Arbitragemöglichkeiten. Wir identifizieren die Abwesenheit von Arbitragemöglichkeiten der ersten Art (NA1) als minimale Eigenschaft, die in jedem finanzmathematischen Modell gelten muss, und zeigen, dass (NA1) äquivalent zur Existenz eines dominierenden lokalen Martingalmaßes ist. Als Beispiel für Prozesse, die (NA1) erfüllen, studieren wir stetige lokale Martingale, die darauf bedingt werden nie Null zu treffen. Anschließend verwenden wir eine modellfreie Version der (NA1) Eigenschaft, die es erlaubt, qualitative Eigenschaften von “typischen Preistrajektorien” zu beschreiben. Hier konstruieren wir ein pfadweises Itô-Integral. Dies deutet an, dass sich typische Preispfade als rough-path-Integratoren verwenden lassen. Nun entwickeln wir mittels Fourierentwicklungen einen alternativen Zugang zur rough-path-Theorie. Wir zerlegen das Integral in drei Operatoren mit verschiedenen Eigenschaften. So wird offensichtlich, dass Integratoren mit der Regularität der Brownschen Bewegung mit ihrer Lévy-Fläche versehen werden müssen, um ein pfadweise stetiges Integral zu erhalten. Daraufhin bemerken wir, dass die Integration zweier Funktionen gegeneinander äquivalent dazu ist, eine Funktion mit der Ableitung einer anderen (im Allgemeinen eine Distribution) zu multiplizieren. In höheren Dimensionen ist das Multiplikationsproblem jedoch allgemeiner. Wir verwenden Littlewood-Paley-Theorie, um unseren Fourier-Zugang zur rough-path-Theorie auf Funktionen mehrdimensionaler Variablen zu erweitern. Wir konstruieren einen Operator, der für Funktionen mit dem punktweisen Produkt übereinstimmt und in einer geeigneten Topologie stetig ist. Nun lassen sich stochastische partielle Differentialgleichungen lösen, die bisher aufgrund von Nichtlinearitäten nicht zugänglich waren. / This thesis deals with various problems from stochastic analysis and from mathematical finance that can best be summarized under the common theme of robustness. We begin by studying financial market models with arbitrage opportunities. We identify the weak notion of absence of arbitrage opportunities of the first kind (NA1) as the minimal property that every sensible asset price model should satisfy, and we prove that (NA1) is equivalent to the existence of a dominating local martingale. As examples of processes that satisfy (NA1) but do not admit equivalent local martingale measures, we study continuous local martingales conditioned not to hit zero. We continue by working with a model free formulation of the (NA1) property, which permits to describe qualitative properties of “typical asset price trajectories”. We construct a pathwise Itô integral for typical price paths. Our results indicate that typical price paths can be used as integrators in the theory of rough paths. Next, we use a Fourier series expansion to develop an alternative approach to rough path integration. We decompose the integral into three components with different behavior. Then it is easy to see that integrators with the regularity of the Brownian motion must be equipped with their Lévy area to obtain a pathwise continuous integral operator. We now note that integrating two functions against each other is equivalent to multiplying one with the derivative of the other, which will in general only be a distribution. In higher index dimensions however, the multiplication problem is more general. We use Littlewood-Paley theory to extend our Fourier approach from rough path integrals to multiplying functions of a multidimensional index. We construct an operator which agrees with the usual product for smooth functions, and which is continuous in a suitable topology. We apply this to solve stochastic partial differential equations that were previously difficult to access due to nonlinearities.

In dieser Arbeit untersuchen wir Superlösungen stochastischer Rückwärtsdifferentialgleichungen (BSDEs) und ein Gleichgewichtsmodell angewandt auf zwei spezifische verallgemeinerte Capital Asset Pricing Models (CAPMs). Unter der Annahme, dass Generatoren der BSDEs unterhalbstetig und von unten durch eine affine Funktion der Kontrollvariablen beschränkt sind sowie eine spezifische Normalisierungseigenschaft erfüllen, beweisen wir Existenz und Eindeutigkeit der minimalen Superlösung, wobei wir Semimartingalkonvergenz und eine geeignet definierte Präorder in Verbindung mit dem Zornschen Lemma nutzen. Anschließend betrachten wir konvexe Generatoren und restringieren admissible Kontrollen auf stetige Semimartingale, wobei wir eine Abhängigkeit des Generators von den Zerlegungsteilen zulassen. Wir beweisen Existenz von Superlösungen, die an endlich vielen Zeitpunkten minimal sind. Neben Stabilitätsresultaten für den nichtlinearen Operator, der einer Endbedingung den Wert der minimalen Superlösung zum Zeitpunk null zuordnet, leiten wir dessen duale Darstellung her und geben eine explizite Form dieser im Falle eines quadratischen Generators an. Ferner geben wir mittels der Dualität Bedingungen für die Existenz von Lösungen unter Nebenbedingungen. Im zweiten Teil der Arbeit behandeln wir ein Gleichgewichtsmodell in stetiger Zeit für verallgemeinerte CAPMs, das endlich viele Agenten und Finanzprodukte umfasst. Die Agenten maximieren exponentielle Nutzenfunktionen und ihre Anfangsausstattung wird von den gehandelten Produkten aufgespannt. Wir zeigen Existenz eines Gleichgewichts, in welchem die optimalen Handelsstrategien konstant sind und von jeweiliger Risikoaversion und Anfangsausstattung abhängen. Hiernach werden affine Prozesse sowie die Theorie des sogenannten Information-based Asset Pricing zur Modellierung herangezogen. Wir leiten semi-explizite Preisformeln her, die sich für effiziente numerische Berechnungen eignen, da keine Monte-Carlo-Methoden gebraucht werden. / In this thesis we study supersolutions of backward stochastic differential equations (BSDEs) and equilibrium pricing within two specific generalized capital asset pricing models (CAPMs). In the first part of the thesis we begin by assuming that the generators of the BSDEs under consideration are jointly lower semicontinuous, bounded from below by an affine function of the control variable, and satisfy a specific normalization property. We prove the existence and uniqueness of the minimal supersolution making use of a particular kind of semimartingale convergence and a suitably defined preorder in combination with Zorn''s lemma. Next, we assume generators to be convex and introduce constraints by restricting admissible controls to continuous semimartingales, where we allow for a dependence of the generator on the respective decomposition parts. We prove existence of supersolutions that are minimal at finitely many fixed times. Besides providing stability results for the non-linear operator that maps a terminal condition to the value of the minimal supersolution at time zero, we give a dual representation of it, including an explicit computation of the conjugate in the case of a quadratic generator, and derive conditions for the existence of solutions under constraints by means of the duality results. In the second part of the thesis we study equilibrium pricing in continuous time within generalized CAPMs. Our model comprises finitely many economic agents and tradable securities. The agents seek to maximize exponential utilities and their endowments are spanned by the securities. We show that an equilibrium exists and the agents'' optimal trading strategies are constant and dependent on their risk aversion and endowment. Affine processes, and the theory of information-based asset pricing are then used for modeling purposes. We derive semi-explicit pricing formulae which lend themselves to efficient numerical computations, as no Monte Carlo methods are needed.

In der vorliegenden Arbeit werden Musterbildungsphänomene auf Submikrometerskalen in reaktiven Adsorbaten auf einkristallinen Katalysatoroberflächen theoretisch untersucht. Da auf solch kleinen Skalen Fluktuationen nicht mehr vernachlässigt werden können, wird eine mesoskopische Theorie entwickelt, die zwischen mikroskopischen Gittermodellen und Reaktions-Diffusions-Systemen vermittelt. Sie beschreibt die Dynamik lokal gemittelter Adsorbatbedeckungen im Rahmen eines Kontinuumsmodells unter Berücksichtigung interner Fluktuationen. Dieser Ansatz wird auf verschiedene Systeme angewendet, in denen sich Muster auf Längenskalen ausbilden, die kleiner als die charakterist ische Diffusionslänge sind, die typischerweise im Mikrometerbereich liegt. Wie beispielsweise in kürzlich durchgefh hrten Experimenten mit einem vergleichsweise schnellen Rastertunnelmikroskop beobachtet wurde, können attraktive Adsorbat-Adsorbat-Wech sel wirkungen zu verschiedenen Mustern auf Nanometerskalen führen. Hier wird zunächst eine einzelne Adsorbatspezies betrachtet. In Abwesenheit von Nichtgleichgewichtsreaktionen können hinreichend starke attraktive laterale Adsorbatwechselwirkungen einen Phasenh bergang erster Ordnung in der Adsorbatbedeckung induzieren. Die mesoskopische Entwicklungsgleichung wird auf die Modellierung der Kinetik dieses Phasenh bergangs angewendet. Berücksichtigt man zusätzlich eine Nichtgleichgewichtsreakti on, so können sich stationäre räumlich periodische Mikrostrukturen aufgrund der Konkurrenz zwischen dem Phasenh bergang und der Reaktion ausbilden. Die Vorraussetzungen für deren Auftreten und ihre charakteristischen Eigenschaften werden hier detailliert analysiert. Unter anderem werden alternierende Wechselwirkungen diskutiert und der Einfluß globaler Kopplung durch die Gasphase auf die Musterbildung wird betrachtet. Außerdem wird gezeigt, da8 die Mikrostrukturen auch durch vergleichsweise starke interne Fluktuationen nicht zerstört werden. Im nächsten Schritt wird ein hypothetisches Modell für zwei verschiedene Adsorbatspezies untersucht, in dem ein ähnlicher Mechanismus zur Bildung von laufenden und stehenden Wellenmustern auf der Nanoskala führt. Werden vergleichsweise starke interne Fluktuationen berücksichtigt, so brechen diese Wellenmuster auf und man beobachtet eine komplexe Dynamik miteinander wechselwirkender Wellenfragmente. Im letzten Beispiel wird anhand der Analyse eines einfachen Modells gezeigt, da8 sich auf Skalen unterhalb der Diffusionslänge selbstorganisierte Mikroreaktoren in einer einzelnen reaktiven Adsorbatspezies ausbilden können, ohne daß die Teilchen miteinander wechselwirken. Sie entsprechen lokalisierten Strukturen, die aufgrund des Zusammenspiels einer Nichtgleichgewichtsreaktion, der Diffusion und eines adsorbatinduzierten strukturellen Phasenh bergangs in der Substratoberfläche entstehen. / Nanoscale pattern formation in reactive adsorbates on single crystal surfaces is investigated theoretically. Because on such small scales fluctuations become important, a mesoscopic theory for the adsorbate coverage is developed, which aims at providing a link between microscopic lattice models and reaction-diffusion equations. It describes the dynamics for the locally averaged adsorbate coverages in a continuum model taking into account internal fluctuations. This approach is applied to several systems, where patterns on scales smaller than the characteristic diffusion length, which typically lies in the micrometer range, can be formed. As has been observed e.g. in recent experiments with fast scanning tunneling microscopy, a variety of nanoscale patterns can result from the presence of attractive adsorbate-adsorbate interactions. Here, at first a single species of such an adsorbate is considered. In the absence of nonequilibrium reactions, strong enough attractive lateral interactions can induce a first-order phase transition in the adsorbate coverage. The mesoscopic evolution equation is applied to model the kinetics of this phase transition. If additionally a nonequilibrium reaction is present, stationary spatially periodic microstructures may arise as a result of the competition of the attractive lateral interactions and the reactions. The conditions for their appearance and their properties are investigated in detail, e.g. alternating lateral interactions are discussed and the influence of global coupling through the gas phase is analyzed. Furthermore, it is shown that they are not destroyed by relatively strong internal fluctuations. In the next step, a hypothetical model for two different reactive adsorbate species is investigated, where a similar mechanism leads to the formation of nanoscale traveling and standing waves. In the presence of relatively strong internal fluctuations these waves break up and a complex dynamics of interacting wave fragments is observed. In the last example, it is shown in the analysis of a simple model that self-organized nonequilibrium microreactors with submicrometer sizes may spontaneously develop in a single reactive adsorbate species without attractive lateral interactions. They represent localized structures resulting from the interplay between reaction, diffusion and an adsorbate-induced structural transformation of the surface.