510 Mathematik
Refine
Year of publication
Document Type
- Doctoral Thesis (75)
- Habilitation (2)
- Article (1)
Keywords
- Optimierung (12)
- Approximationstheorie (7)
- Approximation (6)
- Funktionentheorie (6)
- Partielle Differentialgleichung (6)
- Universalität (6)
- Funktionalanalysis (5)
- universal functions (5)
- Analysis (4)
- Numerische Mathematik (4)
Institute
- Mathematik (63)
- Fachbereich 4 (15)
In this thesis, we study the convergence behavior of an efficient optimization method used for the identification of parameters for underdetermined systems. The research is motivated by optimization problems arising from the estimation of parameters in neural networks as well as in option pricing models. In the first application, we are concerned with neural networks used to forecasting stock market indices. Since neural networks are able to describe extremely complex nonlinear structures they are used to improve the modelling of the nonlinear dependencies occurring in the financial markets. Applying neural networks to the forecasting of economic indicators, we are confronted with a nonlinear least squares problem of large dimension. Furthermore, in this application the number of parameters of the neural network to be determined is usually much larger than the number of patterns which are available for the determination of the unknowns. Hence, the residual function of our least squares problem is underdetermined. In option pricing, an important but usually not known parameter is the volatility of the underlying asset of the option. Assuming that the underlying asset follows a one-factor continuous diffusion model with nonconstant drift and volatility term, the value of an European call option satisfies a parabolic initial value problem with the volatility function appearing in one of the coefficients of the parabolic differential equation. Using this system equation, the estimation of the volatility function is described by a nonlinear least squares problem. Since the adaption of the volatility function is based only on a small number of observed market data these problems are naturally ill-posed. For the solution of these large-scale underdetermined nonlinear least squares problems we use a fully iterative inexact Gauss-Newton algorithm. We show how the structure of a neural network as well as that of the European call price model can be exploited using iterative methods. Moreover, we present theoretical statements for the convergence of the inexact Gauss-Newton algorithm applied to the less examined case of underdetermined nonlinear least squares problems. Finally, we present numerical results for the application of neural networks to the forecasting of stock market indices as well as for the construction of the volatility function in European option pricing models. In case of the latter application, we discretize the parabolic differential equation using a finite difference scheme and we elucidate convergence problems of the discrete scheme when the initial condition is not everywhere differentiable.
This work is concerned with arbitrage bounds for prices of contingent claims under transaction costs, but regardless of other conceivable market frictions. Assumptions on the underlying market are held as weak as convenient for the deduction of meaningful results that make good economic sense. In discrete time we also allow for underlying price processes with uncountable state space. In continuous time the underlying price process is modeled by a semimartingale. For the most part we could avoid any stronger assumptions. The main problems with which we deal in this work are the modelling of (proportional) transaction costs, Fundamental Theorems of Asset Pricing under transaction costs, dual characterizations of arbitrage bounds under transaction costs, Quantile-Hedging under transaction costs, alternatives to the Black-Scholes model in continuous time (under transaction costs). The results apply to stock and currency markets.
Das Konzept der proximalen Mehrschritt-Regularisierung (MSR) auf Folgen von Gittern bei der Lösung inkorrekter Variationsungleichungen wurde von Kaplan und Tichatschke im Jahre 1997 in ihrer Arbeit "Prox-regularization and solution of illposed elliptic variational inequalities" vorgeschlagen und theoretisch motiviert. In demselben Artikel betrachtet man ein allgemeines Problem der partiellen Regularisierung auf einem abgeschlossenen Unterraum. Als Gegenstand der Anwendung solcher Regularisierung können die schlecht gestellten Optimalsteuerprobleme heraustreten, wobei der Unterraum in dem ganzen Prozessraum durch Steuervariablen gebildet wird. Im ersten Kapitel der vorliegenden Dissertation betrachten wir ein abstraktes linear-quadratisches Kontrollproblem in allgemeinen Hilberträumen. Wir diskutieren Voraussetzungen und Bedingungen, unter denen das Kontrollproblem inkorrekt wird. Danach werden zwei allgemeine numerische Verfahren der partiellen Mehrschritt-Regularisierung formuliert. Im ersten Fall untersucht man das MSR-Verfahren, in dem die Zustandsgleichung in einen quadratischen Strafterm eingebettet wird, gemäß der entsprechenden Publikationen von Kaplan und Tichatschke. Im zweiten Fall werden die Ersatzprobleme des MSR-Verfahrens mit exakt erfüllter Zustandsgleichung entwickelt. Im Mittelpunkt sämtlicher Forschungen steht die Konvergenz der approximativen Lösungen von Ersatzproblemen des MSR-Verfahrens gegen ein Element aus der Optimalmenge des Ausgangsproblems. Es stellt sich die Frage: in welchem der genannten Fälle können schwächeren Konvergenzbedingungen für die inneren Approximationen angegeben werden? Um diese Frage aufzuklären, untersuchen wir zwei inkorrekten Kontrollproblme mit elliptischen Zustandsgleichungen und verteilter Steuerung. Das erste Problem kann auf das bekannte Fuller-Problem zurückgeführt werden, für welches eine analytische Lösung mit sogenanntem "chattering regime" existiert und welches ein Basisbeispiel für unsere Aufgaben liefert. Zur Lösung des Fuller-Problems formulieren wir einen MSR-Algorithmus, in dem man mit Fehlern des Strafverfahrens und der FEM-Approximationen rechnen muß. Als Hauptergebnis erhalten wir ein Konvergenzkriterium, das das asymptotische Verhalten von Regularisierungs-, Diskretisierungs- und Strafparametern des MSR-Algorithmus bestimmt. Im letzten Kapitel formulieren wir ein anderes schlecht gestelltes Optimalsteuerproblem mit verteilter Steuerung über dem Polygongebiet. Die Zustandsgleichung wird nun durch ein Poisson-Problem mit gemischten Randbedingungen erzeugt. Solche Aufgabenstellung liefert eine natürliche Erweiterung des auf einer gewöhnlichen Differentialgeichung beruhenden Fuller-Problems auf die Kontrollprobleme mit partiellen Differentialgleichungen. Wir formulieren neuerlich das MSR-Verfahren, in dem man neben dem Diskretisierungsfehler auch einen Berechnungsfehler berücksichtigt. Diesmal verzichten wir aber auf die Straftechniken und stellen die Ersatzprobleme mit exakt erfüllter Zustandsgleichung zusammen. Mit diesem alternativen Zugang und anhand der Falkschen Beweistechniken erhalten wir ein schwächeres und somit auch besseres Konvergenzkriterium für das MSR-Verfahren. Zum Abschluß präsentieren wir Ergebnisse der numerischen Tests, durchgeführt mit dem MSR-Verfahren für ein konkretes Optimalsteuerproblem, dessen Lösung ein zweidimensionales chattering regime aufweist.
The discretization of optimal control problems governed by partial differential equations typically leads to large-scale optimization problems. We consider flow control involving the time-dependent Navier-Stokes equations as state equation which is stamped by exactly this property. In order to avoid the difficulties of dealing with large-scale (discretized) state equations during the optimization process, a reduction of the number of state variables can be achieved by employing a reduced order modelling technique. Using the snapshot proper orthogonal decomposition method, one obtains a low-dimensional model for the computation of an approximate solution to the state equation. In fact, often a small number of POD basis functions suffices to obtain a satisfactory level of accuracy in the reduced order solution. However, the small number of degrees of freedom in a POD based reduced order model also constitutes its main weakness for optimal control purposes. Since a single reduced order model is based on the solution of the Navier-Stokes equations for a specified control, it might be an inadequate model when the control (and consequently also the actual corresponding flow behaviour) is altered, implying that the range of validity of a reduced order model, in general, is limited. Thus, it is likely to meet unreliable reduced order solutions during a control problem solution based on one single reduced order model. In order to get out of this dilemma, we propose to use a trust-region proper orthogonal decomposition (TRPOD) approach. By embedding the POD based reduced order modelling technique into a trust-region framework with general model functions, we obtain a mechanism for updating the reduced order models during the optimization process, enabling the reduced order models to represent the flow dynamics as altered by the control. In fact, a rigorous convergence theory for the TRPOD method is obtained which justifies this procedure also from a theoretical point of view. Benefiting from the trust-region philosophy, the TRPOD method guarantees to save a lot of computational work during the control problem solution, since the original state equation only has to be solved if we intend to update our model function in the trust-region framework. The optimization process itself is completely based on reduced order information only.
This work is concerned with the numerical solution of optimization problems that arise in the context of ground water modeling. Both ground water hydraulic and quality management problems are considered. The considered problems are discretized problems of optimal control that are governed by discretized partial differential equations. Aspects of special interest in this work are inaccurate function evaluations and the ensuing numerical treatment within an optimization algorithm. Methods for noisy functions are appropriate for the considered practical application. Also, block preconditioners are constructed and analyzed that exploit the structure of the underlying linear system. Specifically, KKT systems are considered, and the preconditioners are tested for use within Krylov subspace methods. The project was financed by the foundation Stiftung Rheinland-Pfalz für Innovation and carried out in joint work with TGU GmbH, a company of consulting engineers for ground water and water resources.
The goal of this thesis is to transfer the logarithmic barrier approach, which led to very efficient interior-point methods for convex optimization problems in recent years, to convex semi-infinite programming problems. Based on a reformulation of the constraints into a nondifferentiable form this can be directly done for convex semi- infinite programming problems with nonempty compact sets of optimal solutions. But, by means of an involved max-term this reformulation leads to nondifferentiable barrier problems which can be solved with an extension of a bundle method of Kiwiel. This extension allows to deal with inexact objective values and subgradient information which occur due to the inexact evaluation of the maxima. Nevertheless we are able to prove similar convergence results as for the logarithmic barrier approach in the finite optimization. In the further course of the thesis the logarithmic barrier approach is coupled with the proximal point regularization technique in order to solve ill-posed convex semi-infinite programming problems too. Moreover this coupled algorithm generates sequences converging to an optimal solution of the given semi-infinite problem whereas the pure logarithmic barrier only produces sequences whose accumulation points are such optimal solutions. If there are certain additional conditions fulfilled we are further able to prove convergence rate results up to linear convergence of the iterates. Finally, besides hints for the implementation of the methods we present numerous numerical results for model examples as well as applications in finance and digital filter design.
