This thesis considers classical methods to solve stochastic control problems and valuation problems from financial mathematics numerically. To this end, (linear) partial differential equations (PDEs) in non-divergence form or the optimality conditions known as the (nonlinear) Hamilton-Jacobi-Bellman (HJB) equations are solved by means of finite differences, volumes and elements. We consider all of these three approaches in detail after a thorough introduction to stochastic control problems and discuss various solution terms including classical solutions, strong solutions, weak solutions and viscosity solutions. A particular role in this thesis play degenerate problems. Here, a new model for the optimal control of an energy storage facility is developed which extends the model introduced in [Chen, Forsyth (2007)]. This four-dimensional HJB equation is solved by the classical finite difference Kushner-Dupuis scheme [Kushner, Dupuis (2001)] and a semi-Lagrangian variant which are both discussed in detail. Additionally, a convergence proof of the standard scheme in the setting of parabolic HJB equations is given. Finite volume schemes are another classical method to solve partial differential equations numerically. Sharing similarities to both finite difference and finite element schemes we develop a vertex-centered dual finite volume scheme. We discuss convergence properties and apply the scheme to the solution of HJB equations, which has not been done in such a broad context, to the best of our knowledge. Astonishingly, this is one of the first times the finite volume approach is systematically discussed for the solution of HJB equations. Furthermore, we give many examples which show advantages and disadvantages of the approach. Finally, we investigate novel tailored non-conforming finite element approximations of second-order PDEs in non-divergence form, utilizing finite-element Hessian recovery strategies to approximate second derivatives in the equation. We study approximations with both continuous and discontinuous trial functions. Of particular interest are a-priori and a-posteriori error estimates as well as adaptive finite element methods. In numerical experiments our method is compared with other approaches known from the literature. We discuss implementations of all three approaches in MATLAB (finite differences and volumes) and FEniCS (finite elements) publicly available in GitHub repositories under https://github.com/janblechschmidt. Many numerical experiments show convergence properties as well as pros and cons of the respective approach. Additionally, a new postprocessing procedure for policies obtained from numerical solutions of HJB equations is developed which improves the accuracy of control laws and their incurred values.

The subject of this thesis is the maximization of the information divergence from an exponential family on a finite set, a problem first formulated by Nihat Ay. A special case is the maximization of the mutual information or the multiinformation between different parts of a composite system. My thesis contributes mainly to the mathematical aspects of the optimization problem. A reformulation is found that relates the maximization of the information divergence with the maximization of an entropic quantity, defined on the normal space of the exponential family. This reformulation simplifies calculations in concrete cases and gives theoretical insight about the general problem. A second emphasis of the thesis is on examples that demonstrate how the theoretical results can be applied in particular cases. Third, my thesis contain first results on the characterization of exponential families with a small maximum value of the information divergence.:1. Introduction 2. Exponential families 2.1. Exponential families, the convex support and the moment map 2.2. The closure of an exponential family 2.3. Algebraic exponential families 2.4. Hierarchical models 3. Maximizing the information divergence from an exponential family 3.1. The directional derivatives of D(*|E ) 3.2. Projection points and kernel distributions 3.3. The function DE 3.4. The first order optimality conditions of DE 3.5. The relation between D(*|E) and DE 3.6. Computing the critical points 3.7. Computing the projection points 4. Examples 4.1. Low-dimensional exponential families 4.1.1. Zero-dimensional exponential families 4.1.2. One-dimensional exponential families 4.1.3. One-dimensional exponential families on four states 4.1.4. Other low-dimensional exponential families 4.2. Partition models 4.3. Exponential families with max D(*|E ) = log(2) 4.4. Binary i.i.d. models and binomial models 5. Applications and Outlook 5.1. Principles of learning, complexity measures and constraints 5.2. Optimally approximating exponential families 5.3. Asymptotic behaviour of the empirical information divergence A. Polytopes and oriented matroids A.1. Polytopes A.2. Oriented matroids Bibliography Index Glossary of notations

For ensuring network survivability in case of single component failures many routing protocols provide a primary and a back up routing path for each origin destination pair. We address the problem of selecting these paths such that in the event of multiple failures, occuring with given probabilities, the total loss in routable demand due to both paths being intersected is small with high probability. We present a chance constraint model and solution approaches based on an explicit integer programming formulation, a robust formulation and a cutting plane approach that yield reasonably good solutions assuming that the failures are caused by at most two elementary events, which may each affect several network components.

This thesis addresses optimal control problems with elasticity for large deformations. A hyperelastic model with a polyconvex energy density is employed to describe the elastic behavior of a body. The two approaches to derive the nonlinear partial differential equation, a balance of forces and an energy minimization, are compared. Besides the conventional volume and boundary loads, two novel internal loads are presented. Furthermore, curvilinear coordinates and a hierarchical plate model can be incorporated into the formulation of the elastic forward problem. The forward problem can be solved with Newton\\\'s method, though a globalization technique should be used to avoid divergence of Newton\\\'s method. The repeated solution of the Newton system is done by a CG or MinRes method with a multigrid V-cycle as a preconditioner. The optimal control problem consists of the displacement (as the state) and a load (as the control). Besides the standard tracking-type objective, alternative objective functionals are presented for problems where a reasonable desired state cannot be provided. Two methods are proposed to solve the optimal control problem: an all-at-once approach by a Lagrange-Newton method and a reduced formulation by a quasi-Newton method with an inverse limited-memory BFGS update. The algorithms for the solution of the forward problem and the optimal control problem are implemented in the finite-element software FEniCS, with the geometrical multigrid extension FMG. Numerical experiments are performed to demonstrate the mesh independence of the algorithms and both optimization methods.

In this thesis, a new approach for generating reaction time predictions for Bayesian cognitive models of sequential decision-making is proposed. The method is based on a Markov chain Monte Carlo algorithm that, by utilizing prior distributions and likelihood functions of possible action sequences, generates predictions about the time needed to choose one of these sequences. The plausibility of the reaction time predictions produced by this algorithm was investigated for simple exemplary distributions as well as for prior distributions and likelihood functions of a Bayesian model of habit learning. Simulations showed that the reaction time distributions generated by the Markov chain Monte Carlo sampler exhibit key characteristics of reaction time distributions typically observed in decision-making tasks. The introduced method can be easily applied to various Bayesian models for decision-making tasks with any number of choice alternatives. It thus provides the means to derive reaction time predictions for models where this has not been possible before. / In dieser Arbeit wird ein neuer Ansatz zum Generieren von Reaktionszeitvorhersagen für bayesianische Modelle sequenzieller Entscheidungsprozesse vorgestellt. Der Ansatz basiert auf einem Markov-Chain-Monte-Carlo-Algorithmus, der anhand von gegebenen A-priori-Verteilungen und Likelihood-Funktionen von möglichen Handlungssequenzen Vorhersagen über die Dauer einer Entscheidung für eine dieser Handlungssequenzen erstellt. Die Plausibilität der mit diesem Algorithmus generierten Reaktionszeitvorhersagen wurde für einfache Beispielverteilungen sowie für A-priori-Verteilungen und Likelihood-Funktionen eines bayesianischen Modells zur Beschreibung von Gewohnheitslernen untersucht. Simulationen zeigten, dass die vom Markov-Chain-Monte-Carlo-Sampler erzeugten Reaktionszeitverteilungen charakteristische Eigenschaften von typischen Reaktionszeitverteilungen im Kontext sequenzieller Entscheidungsprozesse aufweisen. Das Verfahren lässt sich problemlos auf verschiedene bayesianische Modelle für Entscheidungsparadigmen mit beliebig vielen Handlungsalternativen anwenden und eröffnet damit die Möglichkeit, Reaktionszeitvorhersagen für Modelle abzuleiten, für die dies bislang nicht möglich war.

In dieser Arbeit werden zwei Klassen von Evolutionsgleichungen in einem Matrixraum-Setting studiert: Hyperbolische Erhaltungsgleichungen und aktive skalare Gleichungen. Für erstere wird untersucht, wann man Oszillationen mit Hilfe polykonvexen Maßen ausschließen kann; für Zweitere wird mit Hilfe von Oszillationen gezeigt, dass es unendlich viele periodische schwache Lösungen gibt.

We describe the large-time moment asymptotics for the parabolic Anderson model where the speed of the diffusion is coupled with time, inducing an acceleration or deceleration. We find a lower critical scale, below which the mass flow gets stuck. On this scale, a new interesting variational problem arises in the description of the asymptotics. Furthermore, we find an upper critical scale above which the potential enters the asymptotics only via some average, but not via its extreme values. We make out altogether five phases, three of which can be described by results that are qualitatively similar to those from the constant-speed parabolic Anderson model in earlier work by various authors. Our proofs consist of adaptations and refinements of their methods, as well as a variational convergence method borrowed from finite elements theory.

This thesis offers a probabilistic solution to robust estimation using a novel adaptive estimator. Reliable state estimation is a mandatory prerequisite for autonomous systems interacting with the real world. The presence of outliers challenges the Gaussian assumption of numerous estimation algorithms, resulting in a potentially skewed estimate that compromises reliability. Many approaches attempt to mitigate erroneous measurements by using a robust loss function – which often comes with a trade-off between robustness and numerical stability. The proposed approach is purely probabilistic and enables adaptive large-scale estimation with non-Gaussian error models. The introduced Adaptive Mixture algorithm combines a nonlinear least squares backend with Gaussian mixtures as the measurement error model. Factor graphs as graphical representations allow an efficient and flexible application to real-world problems, such as simultaneous localization and mapping or satellite navigation. The proposed algorithms are constructed using an approximate expectation-maximization approach, which justifies their design probabilistically. This expectation-maximization is further generalized to enable adaptive estimation with arbitrary probabilistic models. Evaluating the proposed Adaptive Mixture algorithm in simulated and real-world scenarios demonstrates its versatility and robustness. A synthetic range-based localization shows that it provides reliable estimation results, even under extreme outlier ratios. Real-world satellite navigation experiments prove its robustness in harsh urban environments. The evaluation on indoor simultaneous localization and mapping datasets extends these results to typical robotic use cases. The proposed adaptive estimator provides robust and reliable estimation under various instances of non-Gaussian measurement errors.

Lightweight structures became more and more important over the last years. One special class of such structures are short fibre reinforced composites, produced by injection moulding. To avoid expensive experiments for testing the mechanical behaviour of these composites proper material models are needed. Thereby, the stochastic nature of the fibre orientation is the main problem. In this thesis it is looked onto the simulation of such materials in a linear thermoelastic setting. This means the material is described by its heat conduction tensor κ(p), its thermal expansion tensor T(p), and its stiffness tensor C(p). Due to the production process the internal fibre orientation p has to been understood as random variable. As a consequence the previously mentioned material quantities also become random. The classical approach is to average these quantities and solve the linear hermoelastic deformation problem with the averaged expressions. Within this thesis the incorpora- tion of this approach in a time and memory efficient manner in an existing finite element software is shown. Especially for the time and memory efficient improvement several implementation aspects of the underlying software are highlighted. For both - the classical material simulation as well as the time efficient improvement of the software - numerical results are shown. Furthermore, the aforementioned classical approach is extended within this thesis for the simulation of the thermal stresses by using the stochastic nature of the heat conduc tion. This is done by developing it into a series w.r.t. the underlying stochastic. For this series known results from uncertainty quantification are applied. With the help of these results the temperature is developed in a Taylor series. For this Taylor series a suitable expansion point is chosen. Afterwards, this series is incorporated into the computation of the thermal stresses. The advantage of this approach is shown in numerical experiments.

This thesis is concerned with motion planning for the classical two-phase Stefan problem in level set formulation. The interface separating the fluid phases from the solid phases is represented as the zero level set of a continuous function whose evolution is described by the level set equation. Heat conduction in the two phases is modeled by the heat equation. A quadratic tracking-type cost functional that incorporates temperature tracking terms and a control cost term that expresses the desire to have the interface follow a prescribed trajectory by adjusting the heat flux through part of the boundary of the computational domain. The formal Lagrange approach is used to establish a first-order optimality system by applying shape calculus tools. For the numerical solution, the level set equation and its adjoint are discretized in space by discontinuous Galerkin methods that are combined with suitable explicit Runge-Kutta time stepping schemes, while the temperature and its adjoint are approximated in space by the extended finite element method (which accounts for the weak discontinuity of the temperature by a dynamic local modification of the underlying finite element spaces) combined with the implicit Euler method for the temporal discretization. The curvature of the interface which arises in the adjoint system is discretized by a finite element method as well. The projected gradient method, and, in the absence of control constraints, the limited memory BFGS method are used to solve the arising optimization problems. Several numerical examples highlight the potential of the proposed optimal control approach. In particular, they show that it inherits the geometric flexibility of the level set method. Thus, in addition to unidirectional solidification, closed interfaces and changes of topology can be tracked. Finally, the Moreau-Yosida regularization is applied to transform a state constraint on the position of the interface into a penalty term that is added to the cost functional. The optimality conditions for this penalized optimal control problem and its numerical solution are discussed. An example confirms the efficacy of the state constraint. / Die vorliegende Arbeit beschäftigt sich mit einem Optimalsteuerungsproblem für das klassische Stefan-Problem in zwei Phasen. Die Phasengrenze wird als Niveaulinie einer stetigen Funktion modelliert, was die Lösung der so genannten Level-Set-Gleichung erfordert. Durch Anpassen des Wärmeflusses am Rand des betrachteten Gebiets soll ein gewünschter Verlauf der Phasengrenze angesteuert werden. Zusammen mit dem Wunsch, ein vorgegebenes Temperaturprofil zu approximieren, wird dieses Ziel in einem quadratischen Zielfunktional formuliert. Die notwendigen Optimalitätsbedingungen erster Ordnung werden formal mit Hilfe der entsprechenden Lagrange-Funktion und unter Benutzung von Techniken aus der Formoptimierung hergeleitet. Für die numerische Lösung müssen die auftretenden partiellen Differentialgleichungen diskretisiert werden. Dies geschieht im Falle der Level-Set-Gleichung und ihrer Adjungierten auf Basis von unstetigen Galerkin-Verfahren und expliziten Runge-Kutta-Methoden. Die Wärmeleitungsgleichung und die entsprechende Gleichung im adjungierten System werden mit einer erweiterten Finite-Elemente-Methode im Ort sowie dem impliziten Euler-Verfahren in der Zeit diskretisiert. Dieser Zugang umgeht die aufwändige Adaption des Gitters, die normalerweise bei der FE-Diskretisierung von Phasenübergangsproblemen unvermeidbar ist. Auch die Krümmung der Phasengrenze wird numerisch mit Hilfe der Methode der finiten Elemente angenähert. Zur Lösung der auftretenden Optimierungsprobleme werden ein Gradienten-Projektionsverfahren und, im Fall dass keine Kontrollschranken vorliegen, die BFGS-Methode mit beschränktem Speicherbedarf eingesetzt. Numerische Beispiele beleuchten die Stärken des vorgeschlagenen Zugangs. Es stellt sich insbesondere heraus, dass sich die geometrische Flexibilität der Level-Set-Methode auf den vorgeschlagenen Zugang zur optimalen Steuerung vererbt. Zusätzlich zur gerichteten Bewegung einer flachen Phasengrenze können somit auch geschlossene Phasengrenzen sowie topologische Veränderungen angesteuert werden. Exemplarisch, und zwar an Hand einer Beschränkung an die Lage der Phasengrenze, wird auch noch die Behandlung von Zustandsbeschränkungen mittels der Moreau-Yosida-Regularisierung diskutiert. Ein numerisches Beispiel demonstriert die Wirkung der Zustandsbeschränkung.