Ulm, Univ., Diss., 2007.

The paper is devoted to the foundation of approximation methods for integral equations of the form (aI+SbI+K)f=g, where S is the Cauchy singular integral operator on (-1,1) and K is a weakly singular integral operator. Here a,b,g are given functions on (-1,1) and the unknown function f on (-1,1) is looked for. It is assumed that a and b are real-valued and Hölder continuous functions on [-1,1] without common zeros and that g belongs to some weighted space of Hölder continuous functions. In particular, g may have a finite number of singularities. Based on known spectral properties of Cauchy singular integral operators approximation methods for the numerical solution of the above equation are constructed, where both aspects the theoretical convergence and the numerical practicability are taken into account. The weighted uniform convergence of these methods is studied using a general approach based on the theory of approximation spaces. With the help of this approach it is possible to prove simultaneously the stability, the convergence and results on the order of convergence of the approximation methods under consideration.

In this paper, we study the partial Fourier method for treating the Lamé equations in three-­dimensional axisymmetric domains subjected to nonaxisymmetric loads. We consider the mixed boundary value problem of the linear theory of elasticity with the displacement u, the body force f \in (L_2)^3 and homogeneous Dirichlet and Neumann boundary conditions. The partial Fourier decomposition reduces, without any error, the three­dimensional boundary value problem to an infinite sequence of two­dimensional boundary value problems, whose solutions u_n (n = 0,1,2,...) are the Fourier coefficients of u. This process of dimension reduction is described, and appropriate function spaces are given to characterize the reduced problems in two dimensions. The trace properties of these spaces on the rotational axis and some properties of the Fourier coefficients u_n are proved, which are important for further numerical treatment, e.g. by the finite-element method. Moreover, generalized completeness relations are described for the variational equation, the stresses and the strains. The properties of the resulting system of two­dimensional problems are characterized. Particularly, a priori estimates of the Fourier coefficients u_n and of the error of the partial Fourier approximation are given.

Matrix Lyapunov and Riccati equations are an important tool in mathematical systems theory. They are the key ingredients in balancing based model order reduction techniques and linear quadratic regulator problems. For small and moderately sized problems these equations are solved by techniques with at least cubic complexity which prohibits their usage in large scale applications. Around the year 2000 solvers for large scale problems have been introduced. The basic idea there is to compute a low rank decomposition of the quadratic and dense solution matrix and in turn reduce the memory and computational complexity of the algorithms. In this thesis efficiency enhancing techniques for the low rank alternating directions implicit iteration based solution of large scale matrix equations are introduced and discussed. Also the applicability in the context of real world systems is demonstrated. The thesis is structured in seven central chapters. After the introduction chapter 2 introduces the basic concepts and notations needed as fundamental tools for the remainder of the thesis. The next chapter then introduces a collection of test examples spanning from easily scalable academic test systems to badly conditioned technical applications which are used to demonstrate the features of the solvers. Chapter four and five describe the basic solvers and the modifications taken to make them applicable to an even larger class of problems. The following two chapters treat the application of the solvers in the context of model order reduction and linear quadratic optimal control of PDEs. The final chapter then presents the extensive numerical testing undertaken with the solvers proposed in the prior chapters. Some conclusions and an appendix complete the thesis.

In praktischen Aufgabenstellungen können zur Modellierung zufälliger Einflüsse, welche sich durch schwache Abhängigkeiten auszeichnen, schwach korrelierte zufällige Funktionen genutzt werden. Die nähere Untersuchung von Funktionalen schwach korrelierter zufälliger Funktionen ist durch die Gestalt der Lösungen von praktischen Fragestellungen motiviert. Die stochastischen Charakteristiken dieser Lösungen lassen sich im Allgemeinen nicht exakt bestimmen, so dass auf Approximationsverfahren zurückgegriffen werden muss. Diese stehen im Mittelpunkt der Dissertation. Zu Beginn werden Entwicklungen von Momenten und Kumulanten der betrachteten linearen Integralfunktionale schwach korrelierter Prozesse nach der Korrelationslänge des Prozesses hergeleitet und eine Vermutung über die exakte Darstellung der Kumulanten formuliert. Für Integralfunktionale von schwach korrelierten Simulationsprozessen, welche aus der Interpolation von Moving-Average-Prozessen entstehen, werden die definierten Charakteristiken hergeleitet. Außerdem steht die Approximation der unbekannten Dichtefunktion im Fokus der Arbeit. Es werden verschiedene Zugänge genutzt. Eine alternative Herleitung zur bereits in der Literatur untersuchten Gram-Charlier-Entwicklung wird in Form der Edgeworth-Entwicklung angegeben. Des Weiteren werden die Sattelpunkt-Approximation und die Maximum-Entropie-Methode untersucht und anhand von Simulationsergebnissen für Integralfunktionale von Simulationsprozessen miteinander verglichen. / In engineering applications stochastic influences which are characterized by weak dependencies can be modelled, among others, by weakly correlated random functions. The solutions of such problems shape up as integral functionals of weakly correlated random functions which motivates more detailed investigations. In general the exact calculation of stochastic characteristics of such integral functionals is impossible so that we have to be content with approximation methods this thesis focuses on. At the beginning expansions of moments and cumulants of linear integral functionals of weakly correlated random processes with respect to the correlation length are considered and an explicit formula of cumulants is conjectured. For integral functionals of weakly correlated random simulation processes, defined as interpolations of moving average processes, the required expansion coefficients are derived. Furthermore the approximation of the unknown probability density is requested. In the thesis there are different approaches used. First we state an alternative way to achieve the already known Gram Charlier approximation by means of Edgeworth expansion. Then we study two further methods, namely the saddlepoint approximation and the maximum entropy method and compare them on the basis of simulation results for integral functionals of simulation processes.

There are various ways to regularize ill-posed operator equations in Hilbert space. If the underlying operator is accretive then Lavrentiev regularization (singular perturbation) is an immediate choice. The corresponding convergence rates for the regularization error depend on the given smoothness assumptions, and for general accretive operators these may be both with respect to the operator or its adjoint. Previous analysis revealed different convergence rates, and their optimality was unclear, specifically for adjoint source conditions. Based on the fundamental study by T. Kato, Fractional powers of dissipative operators. J. Math. Soc. Japan, 13(3):247--274, 1961, we establish power type convergence rates for this case. By measuring the optimality of such rates in terms on limit orders we exhibit optimality properties of the convergence rates, for general accretive operators under direct and adjoint source conditions, but also for the subclass of nonnegative selfadjoint operators.

The analytical solution of a given inhomogeneous boundary value problem of a linear, elliptic, inhomogeneous partial differential equation and a set of inhomogeneous mixed Dirichlet- and Neumann-type boundary conditions is derived in the present paper. In the context of elasticity theory, the problem arises for a non-conservative symmetric ansatz and an extended constitutive law shown earlier. For convenient user application, the scalar function expressed in cylindrical coordinates is primarily obtained for the general case before being expatiated on a special case of linear boundary conditions.

In vielen Forschungsbereichen wie Medizin, Natur- oder Ingenieurwissenschaften spielt die wissenschaftliche Visualisierung eine wichtige Rolle und hilft Wissenschaftlern neue Erkenntnisse zu gewinnen. Der Hauptgrund hierfür ist, dass Visualisierungen das Unsichtbare sichtbar machen können. So können Visualisierungen beispielsweise den Verlauf von Nervenfasern im Gehirn von Probanden oder den Luftstrom um Hindernisse herum darstellen. Diese Arbeit trägt insbesondere zum Teilgebiet der Strömungsvisualisierung bei, welche sich mit der Untersuchung von Prozessen in Flüssigkeiten und Gasen beschäftigt. Eine beliebte Methode, um Einblicke in komplexe Datensätze zu erhalten, besteht darin, einfache und bekannte Strukturen innerhalb eines Datensatzes aufzuspüren. In der Strömungsvisualisierung führt dies zum Konzept der lokalen Linearisierung und Linearität im Allgemeinen. Dies liegt daran, dass lineare Vektorfelder die einfachste Form von nicht-trivialen Feldern darstellen und diese sehr gut verstanden sind. In der Regel werden simulierte Datensätze in einzelne Zellen diskretisiert, welche auf linearer Interpolation basieren. Beispielsweise können auch stationäre Punkte in der Vektorfeldtopologie mittels linearen Strömungsverhaltens charakterisiert werden. Daher ist Linearität allgegenwärtig. Durch das Verständnis von lokalen linearen Strömungsverhalten in Vektorfeldern konnten verschiedene Visualisierungsmethoden erheblich verbessert werden. Ähnliche Erfolge sind auch für andere Methoden zu erwarten. In dieser Arbeit wird das Konzept der Linearität in der Visualisierung weiterentwickelt. Zunächst wird eine bestehende Definition von linearen Nachbarschaften hin zu affin-linearen Nachbarschaften erweitert. Affin-lineare Nachbarschaften sind Regionen mit einem überwiegend linearem Strömungsverhalten. Es wird eine detaillierte Diskussion über die Definition sowie die gewählten Fehlermaße durchgeführt. Weiterhin wird ein Region Growing-Verfahren vorgestellt, welches affin-lineare Nachbarschaften um beliebige Positionen bis zu einem bestimmten, benutzerdefinierten Fehlerschwellwert extrahiert. Um die lokale Linearität in Vektorfeldern zu messen, wird ein komplementärer Ansatz, welcher die Qualität der bestmöglichen linearen Näherung für eine gegebene n-Ring-Nachbarschaft berechnet, diskutiert. In einer ersten Anwendung werden affin-lineare Nachbarschaften an stationären Punkten verwendet, um deren Einflussbereich sowie ihre Wechselwirkung mit der sie umgebenden, nichtlinearen Strömung, aber auch mit sehr nah benachbarten stationären Punkten zu visualisieren. Insbesondere bei sehr großen Datensätzen kann die analytische Beschreibung der Strömung innerhalb eines linearisierten Bereichs verwendet werden, um Vektorfelder zu komprimieren und vorhandene Visualisierungsansätze zu beschleunigen. Insbesondere sollen eine Reihe von Komprimierungsalgorithmen für gitterbasierte Vektorfelder verbessert werden, welche auf der sukzessiven Entfernung einzelner Gitterkanten basieren. Im Gegensatz zu vorherigen Arbeiten sollen affin-lineare Nachbarschaften als Grundlage für eine Segmentierung verwendet werden, um eine obere Fehlergrenze bereitzustellen und somit eine hohe Qualität der Komprimierungsergebnisse zu gewährleisten. Um verschiedene Komprimierungsansätze zu bewerten, werden die Auswirkungen ihrer jeweiligen Approximationsfehler auf die Stromlinienintegration sowie auf integrationsbasierte Visualisierungsmethoden am Beispiel der numerischen Berechnung von Lyapunov-Exponenten diskutiert. Zum Abschluss dieser Arbeit wird eine mögliche Erweiterung des Linearitätbegriffs für Vektorfelder auf zweidimensionalen Mannigfaltigkeiten vorgestellt, welche auf einer adaptiven, atlasbasierten Vektorfeldzerlegung basiert. / In many research areas, such as medicine, natural sciences or engineering, scientific visualization plays an important role and helps scientists to gain new insights. This is because visualizations can make the invisible visible. For example, visualizations can reveal the course of nerve fibers in the brain of test persons or the air flow around obstacles. This thesis in particular contributes to the subfield of flow visualization, which targets the investigation of processes in fluids and gases. A popular way to gain insights into complex datasets is to identify simple and known structures within a dataset. In case of flow visualization, this leads to the concept of local linearizations and linearity in general. This is because linear vector fields represent the most simple class of non-trivial fields and they are extremely well understood. Typically, simulated datasets are discretized into individual cells that are based on linear interpolation. Also, in vector field topology, stationary points can be characterized by considering the local linear flow behavior in their vicinity. Therefore, linearity is ubiquitous. Through the understanding of local linear flow behavior in vector fields by applying the concept of local linearity, some visualization methods have been improved significantly. Similar successes can be expected for other methods. In this thesis, the use of linearity in visualization is investigated. First, an existing definition of linear neighborhoods is extended towards the affine linear neighborhoods. Affine linear neighborhoods are regions of mostly linear flow behavior. A detailed discussion of the definition and of the chosen error measures is provided. Also a region growing algorithm that extracts affine linear neighborhoods around arbitrary positions up to a certain user-defined approximation error threshold is introduced. To measure the local linearity in vector fields, a complementary approach that computes the quality of the best possible linear approximation for a given n-ring neighborhood is discussed. As a first application, the affine linear neighborhoods around stationary points are used to visualize their region of influence, their interaction with the non-linear flow around them as well as their interaction with closely neighbored stationary points. The analytic description of the flow within a linearized region can be used to compress vector fields and accelerate existing visualization approaches, especially in case of very large datasets. In particular, the presented method aims at improving over a series of compression algorithms for grid-based vector fields that are based on edge collapse. In contrast to previous approaches, affine linear neighborhoods serve as the basis for a segmentation in order to provide an upper error bound and also to ensure a high quality of the compression results. To evaluate different compression approaches, the impact of their particular approximation errors on streamline integration as well as on integration-based visualization methods is discussed on the example of Finite-Time Lyapunov Exponent computations. To conclude the thesis, a first possible extension of linearity to fields on two-dimensional manifolds, based on an adaptive atlas-based vector field decomposition, is given.

In this thesis, we address the identification of biomarkers in high-dimensional omics data. The identification of valid biomarkers is especially relevant for personalized medicine that depends on accurate prediction rules. Moreover, biomarkers elucidate the provenance of disease, or molecular changes related to disease. From a statistical point of view the identification of biomarkers is best cast as variable selection. In particular, we refer to variables as the molecular attributes under investigation, e.g. genes, genetic variation, or metabolites; and we refer to observations as the specific samples whose attributes we investigate, e.g. patients and controls. Variable selection in high-dimensional omics data is a complicated challenge due to the characteristic structure of omics data. For one, omics data is high-dimensional, comprising cellular information in unprecedented details. Moreover, there is an intricate correlation structure among the variables due to e.g internal cellular regulation, or external, latent factors. Variable selection for uncorrelated data is well established. In contrast, there is no consensus on how to approach variable selection under correlation. Here, we introduce a multivariate framework for variable selection that explicitly accounts for the correlation among markers. In particular, we present two novel quantities for variable importance: the correlation-adjusted t (CAT) score for classification, and the correlation-adjusted (marginal) correlation (CAR) score for regression. The CAT score is defined as the Mahalanobis-decorrelated t-score vector, and the CAR score as the Mahalanobis-decorrelated correlation between the predictor variables and the outcome. We derive the CAT and CAR score from a predictive point of view in linear discriminant analysis and regression; both quantities assess the weight of a decorrelated and standardized variable on the prediction rule. Furthermore, we discuss properties of both scores and relations to established quantities. Above all, the CAT score decomposes Hotelling’s T 2 and the CAR score the proportion of variance explained. Notably, the decomposition of total variance into explained and unexplained variance in the linear model can be rewritten in terms of CAR scores. To render our approach applicable on high-dimensional omics data we devise an efficient algorithm for shrinkage estimates of the CAT and CAR score. Subsequently, we conduct extensive simulation studies to investigate the performance of our novel approaches in ranking and prediction under correlation. Here, CAT and CAR scores consistently improve over marginal approaches in terms of more true positives selected and a lower model error. Finally, we illustrate the application of CAT and CAR score on real omics data. In particular, we analyze genomics, transcriptomics, and metabolomics data. We ascertain that CAT and CAR score are competitive or outperform state of the art techniques in terms of true positives detected and prediction error.

In mathematics the parameter is used as a special kind of a variable. The classification of the terms \"variable\" and \"parameter\" is often done by intuition and changes due to different situations and needs. The history of mathematics shows that these two terms represent the same abstract object in mathematics. In today´s mathematics, compared to variables, the parameter is declared as an unknown constant measure. This interpretation of parameters can be used in set theory for describing sets with an infinite number of elements. Due to this perspective the structure of vector spaces can be developed as a special structured set theory. Further, the concept of parameters can be seen as a model for developing mathematics education in linear algebra.