Library

Description: In optimal control problems with nonlinear time-dependent 3D PDEs, the computation of the reduced gradient by adjoint methods requires one solve of the state equation forward in time, and one backward solve of the adjoint equation. Since the state enters into the adjoint equation, the storage of a 4D discretization is necessary. We propose a lossy compression algorithm using a cheap predictor for the state data, with additional entropy coding of prediction errors. Analytical and numerical results indicate that compression factors around 30 can be obtained without exceeding the FE discretization error.

Description: We consider an optimal control problem from hyperthermia treatment planning and its barrier regularization. We derive basic results, which lay the groundwork for the computation of optimal solutions via an interior point path-following method. Further, we report on a numerical implementation of such a method and its performance at an example problem.

Description: Pulse thermography of concrete structures is used in civil engineering for detecting voids, honeycombing and delamination. The physical situation is readily modeled by Fourier's law. Despite the simplicity of the PDE structure, quantitatively realistic numerical 3D simulation faces two major obstacles. First, the short heating pulse induces a thin boundary layer at the heated surface which encapsulates all information and therefore has to be resolved faithfully. Even with adaptive mesh refinement techniques, obtaining useful accuracies requires an unsatisfactorily fine discretization. Second, bulk material parameters and boundary conditions are barely known exactly. We address both issues by a semi-analytic reformulation of the heat transport problem and by parameter identification. Numerical results are compared with measurements of test specimens.

Description: Reasons for the failure of adaptive methods to deliver improved efficiency when integrating monodomain models for myocardiac excitation are discussed. Two closely related techniques for reducing the computational complexity of linearly implicit integrators, deliberate sparsing and splitting, are investigated with respect to their impact on computing time and accuracy.

Description: The biodomain model of cardioelectric excitation consists of a reaction‐diffusion equation, an elliptic algebraic constraint, and a set of pointwise ODEs. Fast reaction enforces small time steps, such that for common mesh sizes the reaction‐diffusion equation is easily solved implicitly due to a dominating mass matrix. In contrast, the elliptic constraint does not benefit from small time steps and requires a comparably expensive solution. We propose a delayed residual compensation that improves the solution of the elliptic constraint and thus alleviates the need for long iteration times.

Description: We consider a system with Poisson arrivals and i.i.d. service times and where the requests are served according to the state-dependent (Cohen's generalized) processor sharing discipline, where each request in the system receives a service capacity which depends on the actual number of requests in the system. For this system we derive asymptotically tight upper bounds for the moments of the conditional sojourn time of a request with given required service time. The bounds generalize corresponding results, recently given for the single-server processor sharing system by Cheung et al. and for the state-dependent processor sharing system with exponential service times by the authors. Analogous results hold for the waiting times.

Description: The goal of data fusion is to combine several representations of one real world object into a single, consistent representation, e.g., in data integration. A very popular operator to perform data fusion is the minimum union operator. It is defined as the outer union and the subsequent removal of subsumed tuples. Minimum union is used in other applications as well, for instance in database query optimization to rewrite outer join queries, in the semantic web community in implementing Sparql's optional operator, etc. Despite its wide applicability, there are only few efficient implementations, and until now, minimum union is not a relational database primitive. This paper fills this gap as we present implementations of subsumption that serve as a building block for minimum union. Furthermore, we consider this operator as database primitive and show how to perform optimization of query plans in presence of subsumption and minimum union through rule-based plan transformations. Experiments on both artificial and real world data show that our algorithms outperform existing algorithms used for subsumption in terms of runtime and they scale to large volumes of data. In the context of data integration, we observe that performing data fusion calls for more than subsumption and minimum union. Therefore, another contribution of this paper is the definition of the complementation and complement union operators. Intuitively, these allow to merge tuples that have complementing values and thus eliminate unnecessary null-values.

Description: A data integration process consists of mapping source data into a target representation (schema mapping), identifying multiple representations of the same real-word object (duplicate detection), and finally combining these representations into a single consistent representation (data fusion). Clearly, as multiple representations of an object are generally not exactly equal, during data fusion, we have to take special care in handling data conflicts. This paper focuses on the definition and implementation of complement union, an operator that defines a new semantics for data fusion.

Description: We present Undercover, a primal heuristic for mixed-integer nonlinear programming (MINLP). The heuristic constructs a mixed-integer linear subproblem (sub-MIP) of a given MINLP by fixing a subset of the variables. We solve a set covering problem to identify a minimal set of variables which need to be fixed in order to linearise each constraint. Subsequently, these variables are fixed to approximate values, e.g. obtained from a linear outer approximation. The resulting sub-MIP is solved by a mixed-integer linear programming solver. Each feasible solution of the sub-MIP corresponds to a feasible solution of the original problem. Although general in nature, the heuristic seems most promising for mixed-integer quadratically constrained programmes (MIQCPs). We present computational results on a general test set of MIQCPs selected from the MINLPLib.

Description: In this thesis, we present the theoretical background, implementational details and computational results concerning the generic branch-cut-and-price solver GCG.

Description: We propose a hybrid approach for solving the resource-constrained project scheduling problem which is an extremely hard to solve combinatorial optimization problem of practical relevance. Jobs have to be scheduled on (renewable) resources subject to precedence constraints such that the resource capacities are never exceeded and the latest completion time of all jobs is minimized. The problem has challenged researchers from different communities, such as integer programming (IP), constraint programming (CP), and satisfiability testing (SAT). Still, there are instances with 60 jobs which have not been solved for many years. The currently best known approach, lazyFD, is a hybrid between CP and SAT techniques. In this paper we propose an even stronger hybridization by integrating all the three areas, IP, CP, and SAT, into a single branch-and-bound scheme. We show that lower bounds from the linear relaxation of the IP formulation and conflict analysis are key ingredients for pruning the search tree. First computational experiments show very promising results. For five instances of the well-known PSPLIB we report an improvement of lower bounds. Our implementation is generic, thus it can be potentially applied to similar problems as well.

Description: This paper reports on the fifth version of the Mixed Integer Programming Library. The MIPLIB 2010 is the first MIPLIB release that has been assembled by a large group from academia and from industry, all of whom work in integer programming. There was mutual consent that the concept of the library had to be expanded in order to fulfill the needs of the community. The new version comprises 361 instances sorted into several groups. This includes the main benchmark test set of 87 instances, which are all solvable by today's codes, and also the challenge test set with 164 instances, many of which are currently unsolved. For the first time, we include scripts to run automated tests in a predefined way. Further, there is a solution checker to test the accuracy of provided solutions using exact arithmetic.

Description: Telecommunication is fundamental for the information society. In both, the private and the professional sector, mobile communication is nowadays taken for granted. Starting primarily as a service for speech communication, data service and mobile Internet access are now driving the evolution of network infrastructure. In the year 2009, 19 million users generated over 33 million GB of traffic using mobile data services. The 3rd generation networks (3G or UMTS) in Germany comprises over 39,000 base stations with some 120,000 cells. From 1998 to 2008, the four network operators in Germany invested over 33 billion Euros in their infrastructure. A careful allocation of the resources is thus crucial for the profitability for a network operator: a network should be dimensioned to match customers demand. As this demand evolves over time, the infrastructure has to evolve accordingly. The demand evolution is hard to predict and thus constitutes a strong source of uncertainty. Strategic network planning has to take this uncertainty into account, and the planned network evolution should adapt to changing market conditions. The application of superior planning methods under the consideration of uncertainty can improve the profitability of the network and creates a competitive advantage. Multistage stochastic programming is a suitable framework to model strategic telecommunication network planning. We present mathematical models and effective optimization procedures for strategic cellular network design. The demand evolution is modeled as a continuous stochastic process which is approximated by a discrete scenario tree. A tree-stage approach is used for the construction of non-uniform scenario trees that serve as input of the stochastic program. The model is calibrated by historical traffic observations. A realistic system model of UMTS radio cells is used that determines coverage areas and cell capacities and takes signal propagation and interferences into account. The network design problem is formulated as a multistage stochastic mixed integer linear program, which is solved using state-of-the-art commercial MIP solvers. Problem specific presolving is proposed to reduce the problem size. Computational results on realistic data is presented. Optimization for the expected profit and the conditional value at risk are performed and compared.

Description: In this work, we study the spectra and eigenmodes of the Hessian of various discrete surface energies and discuss applications to shape analysis. In particular, we consider a physical model that describes the vibration modes and frequencies of a surface through the eigenfunctions and eigenvalues of the Hessian of a deformation energy, and we derive a closed form representation for the Hessian (at the rest state of the energy) for a general class of deformation energies. Furthermore, we design a quadratic energy, such that the eigenmodes of the Hessian of this energy are sensitive to the extrinsic curvature of the surface. Based on these spectra and eigenmodes, we derive two shape signatures. One that measures the similarity of points on a surface, and another that can be used to identify features of the surface. In addition, we discuss a spectral quadrangulation scheme for surfaces.

Description: We present a visual approach for the exploration of stress tensor fields. Therefore, we introduce the idea of multiple linked views to tensor visualization. In contrast to common tensor visualization methods that only provide a single view to the tensor field, we pursue the idea of providing various perspectives onto the data in attribute and object space. Especially in the context of stress tensors, advanced tensor visualization methods have a young tradition. Thus, we propose a combination of visualization techniques domain experts are used to with statistical views of tensor attributes. The application of this concept to tensor fields was achieved by extending the notion of shape space. It provides an intuitive way of finding tensor invariants that represent relevant physical properties. Using brushing techniques, the user can select features in attribute space, which are mapped to displayable entities in a three-dimensional hybrid visualization in object space. Volume rendering serves as context, while glyphs encode the whole tensor information in focus regions. Tensorlines can be included to emphasize directionally coherent features in the tensor field. We show that the benefit of such a multi-perspective approach is manifold. Foremost, it provides easy access to the complexity of tensor data. Moreover, including wellknown analysis tools, such as Mohr diagrams, users can familiarize themselves gradually with novel visualization methods. Finally, by employing a focus-driven hybrid rendering, we significantly reduce clutter, which was a major problem of other three-dimensional tensor visualization methods.

Description: This thesis reports on simulating spin models on Nvidia graphics cards using the CUDA programming model; a particular approach for making GPGPU (General Purpose Computation on Graphics Processing Units) available for a wide range of software developers not necessarily acquainted with (massively) parallel programming. By comparing program execution times for simulations of the Ising model and the Ising spin glass by means of the Metropolis algorithm on Nvidia Tesla C1060 graphics cards and an Intel Core i7-920 quad-core x86 CPU (we used OpenMP to make our simulations run on all 4 execution units of the CPU), we noticed that the Tesla C1060 performed about a factor 5-10 faster than the Core i7-920, depending on the particular model and the accuracy of the calculations (32-bit or 64-bit). We also investigated the reliability of GPGPU computations, especially with respect to the occurrence of soft-errors as suggested in [23]. We noticed faulty program outputs during long-time simulations of the Ising model on ''large'' lattices. We were able to link these problems to overheating of the corresponding graphics cards. Doing Monte Carlo simulations on parallel computer architectures, as was the case in this thesis, suggests to also generate random numbers in a parallel manner. We present implementations of the random number generators Ranlux and Mersenne Twister. In addition, we give an alternative and very efficient approach for producing parallel random numbers on Nvidia graphics cards. We successfully tested all random number generators used in this thesis for their quality by comparing Monte Carlo estimates against exact calculations.

Description: Vorliegende Diplomarbeit befasst sich mit der Simulation von Spin-Modellen auf Nvidia Grafikkarten. Hierbei wird das Programmiermodell CUDA verwendet, welches einer breiten Masse von Softwareentwicklern den Zugang zu GPGPU (General Purpose Computation on Graphics Processing Units) gestattet ohne dass diese notwendigerweise mit (massiv) paralleler Programmierung vertraut sein müssen. Im Rahmen von Vergleichen der Programmlaufzeiten für Simulationen des Ising- Modells sowie des Ising-Spin-Glases auf Nvidia Tesla C1060 Grafikkarten und einer Intel Core i7-920 x86 Quad-Core-CPU war zu vermerken, dass, abhängig vom konkreten Modell und der Rechengenauigkeit (32-bit oder 64-bit), die Tesla C1060 zwischen 5-10 mal schneller arbeitete als die Core i7-920 Quad-Core-CPU. Wir haben uns ebenfalls mit der Zuverlässigkeit von GPGPU-Rechnungen befasst, gerade in Hinblick auf das Auftreten von Soft-Errors, wie es in [23] angedeutet wird. Wir beobachteten falsche Programmausgaben bei langen Simulationen des Ising-Modells auf ,,großen'''' Gittern. Es gelang uns die auftretenden Probleme mit Überhitzung der entsprechenden Grafikkarten in Verbindung zu bringen. Für die Durchführung von Monte-Carlo-Simulationen auf Parallelrechnerarchitekturen, wie es in vorliegender Arbeit der Fall ist, liegt es nahe auch Zufallszahlen parallel zu erzeugen. Wir präsentieren Implementierungen der Zufallszahlengeneratoren Ranlux und Mersenne Twister. Zusätzlich stellen wir eine alternative und sehr effiziente Möglichkeit vor parallele Zufallszahlen auf Nvidia Grafikkarten zu erzeugen. Alle verwendeten Zufallszahlengeneratoren wurden erfolgreich auf ihre Qualität getestet indem Monte-Carlo-Schätzer exakten Rechnungen gegenübergestellt wurden.

Description: In contrast to the well known meshbased methods like the finite element method, meshfree methods do not rely on a mesh. However besides their great applicability, meshfree methods are rather time consuming. Thus, it seems favorable to combine both methods, by using meshfree methods only in a small part of the domain, where a mesh is disadvantageous, and a meshbased method for the rest of the domain. We motivate, that this coupling between the two simulation techniques can be considered as saddle point problem and show the stability of this coupling. Thereby a novel transfer operator is introduced, which interacts in the transition zone, where both methods coexist.

Description: This paper is concerned with the design, analysis, and implementation of concurrent coupling approaches where different (atomic and continuous) models are used simultaneously within a single simulation process. Thereby, several problems or pitfalls can happen, for example, the reflection of molecular movements at the “boundary” between the atomic and continuum regions which leads to an unphysical increase in energy in the atomic model. We investigate the problems with the aim of giving an introduction into this field and preventing errors for scientists starting their research towards multiscale methods.