Refine
Year of publication
Document Type
- Doctoral Thesis (44)
- Habilitation (2)
- Article (1)
Language
- English (47) (remove)
Keywords
- Optimierung (6)
- Funktionalanalysis (5)
- Partielle Differentialgleichung (5)
- Approximation (4)
- Numerische Strömungssimulation (4)
- Shape Optimization (4)
- Approximationstheorie (3)
- Funktionentheorie (3)
- Hadamard product (3)
- Kompositionsoperator (3)
Institute
- Mathematik (47) (remove)
In a paper of 1996 the british mathematician Graham R. Allan posed the question, whether the product of two stable elements is again stable. Here stability describes the solvability of a certain infinite system of equations. Using a method from the theory of homological algebra, it is proved that in the case of topological algebras with multiplicative webs, and thus in all common locally convex topological algebras that occur in standard analysis, the answer of Allan's question is affirmative.
In splitting theory of locally convex spaces we investigate evaluable characterizations of the pairs (E, X) of locally convex spaces such that each exact sequence 0 -> X -> G -> E -> 0 of locally convex spaces splits, i.e. either X -> G has a continuous linear left inverse or G -> E has a continuous linear right inverse. In the thesis at hand we deal with splitting of short exact sequences of so-called PLH spaces, which are defined as projective limits of strongly reduced spectra of strong duals of Fréchet-Hilbert spaces. This class of locally convex spaces contains most of the spaces of interest for application in the theory of partial differential operators as the space of Schwartz distributions , the space of real analytic functions and various spaces of ultradifferentiable functions and ultradistributions. It also contains non-Schwartz spaces as B(2,k,loc)(Ω) and spaces of smooth and square integrable functions that are not covered by the current theory for PLS spaces. We prove a complete characterizations of the above problem in the case of X being a PLH space and E either being a Fréchet-Hilbert space or a strong dual of one by conditions of type (T ). To this end, we establish the full homological toolbox of Yoneda Ext functors in exact categories for the category of PLH spaces including the long exact sequence, which in particular involves a thorough discussion of the proper concept of exactness. Furthermore, we exhibit the connection to the parameter dependence problem via the Hilbert tensor product for hilbertizable locally convex spaces. We show that the Hilbert tensor product of two PLH spaces is again a PLH space which in particular proves the positive answer to Grothendieck- problème des topologies. In addition to that we give a complete characterization of the vanishing of the first derivative of the functor proj for tensorized PLH spectra if one of the PLH spaces E and X meets some nuclearity assumptions. To apply our results to concrete cases we establish sufficient conditions of (DN)-(Ω) type and apply them to the parameter dependence problem for partial differential operators with constant coefficients on B(2,k,loc)(Ω) spaces as well as to the smooth and square integrable parameter dependence problem. Concluding we give a complete solution of all the problems under consideration for PLH spaces of Köthe type.
Krylov subspace methods are often used to solve large-scale linear equations arising from optimization problems involving partial differential equations (PDEs). Appropriate preconditioning is vital for designing efficient iterative solvers of this type. This research consists of two parts. In the first part, we compare two different kinds of preconditioners for a conjugate gradient (CG) solver attacking one partial integro-differential equation (PIDE) in finance, both theoretically and numerically. An analysis on mesh independence and rate of convergence of the CG solver is included. The knowledge of preconditioning the PIDE is applied to a relevant optimization problem. The second part aims at developing a new preconditioning technique by embedding reduced order models of nonlinear PDEs, which are generated by proper orthogonal decomposition (POD), into deflated Krylov subspace algorithms in solving corresponding optimization problems. Numerical results are reported for a series of test problems.
Copositive programming is concerned with the problem of optimizing a linear function over the copositive cone, or its dual, the completely positive cone. It is an active field of research and has received a growing amount of attention in recent years. This is because many combinatorial as well as quadratic problems can be formulated as copositive optimization problems. The complexity of these problems is then moved entirely to the cone constraint, showing that general copositive programs are hard to solve. A better understanding of the copositive and the completely positive cone can therefore help in solving (certain classes of) quadratic problems. In this thesis, several aspects of copositive programming are considered. We start by studying the problem of computing the projection of a given matrix onto the copositive and the completely positive cone. These projections can be used to compute factorizations of completely positive matrices. As a second application, we use them to construct cutting planes to separate a matrix from the completely positive cone. Besides the cuts based on copositive projections, we will study another approach to separate a triangle-free doubly nonnegative matrix from the completely positive cone. A special focus is on copositive and completely positive programs that arise as reformulations of quadratic optimization problems. Among those we start by studying the standard quadratic optimization problem. We will show that for several classes of objective functions, the relaxation resulting from replacing the copositive or the completely positive cone in the conic reformulation by a tractable cone is exact. Based on these results, we develop two algorithms for solving standard quadratic optimization problems and discuss numerical results. The methods presented cannot immediately be adapted to general quadratic optimization problems. This is illustrated with examples.
In this thesis, we aim to study the sampling allocation problem of survey statistics under uncertainty. We know that the stratum specific variances are generally not known precisely and we have no information about the distribution of uncertainty. The cost of interviewing each person in a stratum is also a highly uncertain parameter as sometimes people are unavailable for the interview. We propose robust allocations to deal with the uncertainty in both stratum specific variances and costs. However, in real life situations, we can face such cases when only one of the variances or costs is uncertain. So we propose three different robust formulations representing these different cases. To the best of our knowledge robust allocation in the sampling allocation problem has not been considered so far in any research.
The first robust formulation for linear problems was proposed by Soyster (1973). Bertsimas and Sim (2004) proposed a less conservative robust formulation for linear problems. We study these formulations and extend them for the nonlinear sampling allocation problem. It is very unlikely to happen that all of the stratum specific variances and costs are uncertain. So the robust formulations are in such a way that we can select how many strata are uncertain which we refer to as the level of uncertainty. We prove that an upper bound on the probability of violation of the nonlinear constraints can be calculated before solving the robust optimization problem. We consider various kinds of datasets and compute robust allocations. We perform multiple experiments to check the quality of the robust allocations and compare them with the existing allocation techniques.
Considering the numerical simulation of mathematical models it is necessary to have efficient methods for computing special functions. We will focus our considerations in particular on the classes of Mittag-Leffler and confluent hypergeometric functions. The PhD Thesis can be structured in three parts. In the first part, entire functions are considered. If we look at the partial sums of the Taylor series with respect to the origin we find that they typically only provide a reasonable approximation of the function in a small neighborhood of the origin. The main disadvantages of these partial sums are the cancellation errors which occur when computing in fixed precision arithmetic outside this neighborhood. Therefore, our aim is to quantify and then to reduce this cancellation effect. In the next part we consider the Mittag-Leffler and the confluent hypergeometric functions in detail. Using the method we developed in the first part, we can reduce the cancellation problems by "modifying" the functions for several parts of the complex plane. Finally, in in the last part two other approaches to compute Mittag-Leffler type and confluent hypergeometric functions are discussed. If we want to evaluate such functions on unbounded intervals or sectors in the complex plane, we have to consider methods like asymptotic expansions or continued fractions for large arguments z in modulus.
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.
Optimal Control of Partial Integro-Differential Equations and Analysis of the Gaussian Kernel
(2018)
An important field of applied mathematics is the simulation of complex financial, mechanical, chemical, physical or medical processes with mathematical models. In addition to the pure modeling of the processes, the simultaneous optimization of an objective function by changing the model parameters is often the actual goal. Models in fields such as finance, biology or medicine benefit from this optimization step.
While many processes can be modeled using an ordinary differential equation (ODE), partial differential equations (PDEs) are needed to optimize heat conduction and flow characteristics, spreading of tumor cells in tissue as well as option prices. A partial integro-differential equation (PIDE) is a parital differential equation involving an integral operator, e.g., the convolution of the unknown function with a given kernel function. PIDEs occur for example in models that simulate adhesive forces between cells or option prices with jumps.
In each of the two parts of this thesis, a certain PIDE is the main object of interest. In the first part, we study a semilinear PIDE-constrained optimal control problem with the aim to derive necessary optimality conditions. In the second, we analyze a linear PIDE that includes the convolution of the unknown function with the Gaussian kernel.
Recently, optimization has become an integral part of the aerodynamic design process chain. However, because of uncertainties with respect to the flight conditions and geometrical uncertainties, a design optimized by a traditional design optimization method seeking only optimality may not achieve its expected performance. Robust optimization deals with optimal designs, which are robust with respect to small (or even large) perturbations of the optimization setpoint conditions. The resulting optimization tasks become much more complex than the usual single setpoint case, so that efficient and fast algorithms need to be developed in order to identify, quantize and include the uncertainties in the overall optimization procedure. In this thesis, a novel approach towards stochastic distributed aleatory uncertainties for the specific application of optimal aerodynamic design under uncertainties is presented. In order to include the uncertainties in the optimization, robust formulations of the general aerodynamic design optimization problem based on probabilistic models of the uncertainties are discussed. Three classes of formulations, the worst-case, the chance-constrained and the semi-infinite formulation, of the aerodynamic shape optimization problem are identified. Since the worst-case formulation may lead to overly conservative designs, the focus of this thesis is on the chance-constrained and semi-infinite formulation. A key issue is then to propagate the input uncertainties through the systems to obtain statistics of quantities of interest, which are used as a measure of robustness in both robust counterparts of the deterministic optimization problem. Due to the highly nonlinear underlying design problem, uncertainty quantification methods are used in order to approximate and consequently simplify the problem to a solvable optimization task. Computationally demanding evaluations of high dimensional integrals resulting from the direct approximation of statistics as well as from uncertainty quantification approximations arise. To overcome the curse of dimensionality, sparse grid methods in combination with adaptive refinement strategies are applied. The reduction of the number of discretization points is an important issue in the context of robust design, since the computational effort of the numerical quadrature comes up in every iteration of the optimization algorithm. In order to efficiently solve the resulting optimization problems, algorithmic approaches based on multiple-setpoint ideas in combination with one-shot methods are presented. A parallelization approach is provided to overcome the amount of additional computational effort involved by multiple-setpoint optimization problems. Finally, the developed methods are applied to 2D and 3D Euler and Navier-Stokes test cases verifying their industrial usability and reliability. Numerical results of robust aerodynamic shape optimization under uncertain flight conditions as well as geometrical uncertainties are presented. Further, uncertainty quantification methods are used to investigate the influence of geometrical uncertainties on quantities of interest in a 3D test case. The results demonstrate the significant effect of uncertainties in the context of aerodynamic design and thus the need for robust design to ensure a good performance in real life conditions. The thesis proposes a general framework for robust aerodynamic design attacking the additional computational complexity of the treatment of uncertainties, thus making robust design in this sense possible.
The main achievement of this thesis is an analysis of the accuracy of computations with Loader's algorithm for the binomial density. This analysis in later progress of work could be used for a theorem about the numerical accuracy of algorithms that compute rectangle probabilities for scan statistics of a multinomially distributed random variable. An example that shall illustrate the practical use of probabilities for scan statistics is the following, which arises in epidemiology: Let n patients arrive at a clinic in d = 365 days, each of the patients with probability 1/d at each of these d days and all patients independently from each other. The knowledge of the probability, that there exist 3 adjacent days, in which together more than k patients arrive, helps deciding, after observing data, if there is a cluster which we would not suspect to have occurred randomly but for which we suspect there must be a reason. Formally, this epidemiological example can be described by a multinomial model. As multinomially distributed random variables are examples of Markov increments, which is a fact already used implicitly by Corrado (2011) to compute the distribution function of the multinomial maximum, we can use a generalized version of Corrado's Algorithm to compute the probability described in our example. To compute its result, the algorithm for rectangle probabilities for Markov increments always uses transition probabilities of the corresponding Markov Chain. In the multinomial case, the transition probabilities of the corresponding Markov Chain are binomial probabilities. Therefore, we start an analysis of accuracy of Loader's algorithm for the binomial density, which for example the statistical software R uses. With the help of accuracy bounds for the binomial density we would be able to derive accuracy bounds for the computation of rectangle probabilities for scan statistics of multinomially distributed random variables. To figure out how sharp derived accuracy bounds are, in examples these can be compared to rigorous upper bounds and rigorous lower bounds which we obtain by interval-arithmetical computations.