>Magister Scientiae - MSc / In Monte Carlo path simulations, which are used extensively in computational -finance, one is interested in the expected value of a quantity which is a functional of the solution to a stochastic differential equation [M.B. Giles, Multilevel Monte Carlo Path Simulation: Operations Research, 56(3) (2008) 607-617] where we have a scalar function with a uniform Lipschitz bound. Normally, we discretise the stochastic differential equation numerically. The simplest estimate for this expected value is the mean of the payoff (the value of an option at the terminal period) values from N independent path simulations. The multilevel Monte Carlo path simulation method recently introduced by Giles exploits strong convergence properties to improve the computational complexity by combining simulations with different levels of resolution. This new method improves on the computational complexity of the standard Monte Carlo approach by considering Monte Carlo simulations with a geometric sequence of different time steps following the approach of Kebaier [A. Kebaier, Statistical Romberg extrapolation: A new variance reduction method and applications to options pricing. Annals of Applied Probability 14(4) (2005) 2681- 2705]. The multilevel method makes computation easy as it estimates each of the terms of the estimate independently (as opposed to the Monte Carlo method) such that the computational complexity of Monte Carlo path simulations is minimised. In this thesis, we investigate this method in pricing path-dependent options and the computation of option price sensitivities also known as Greeks.

Title: Analysis of Krylov subspace methods Author: Tomáš Gergelits Department: Department of Numerical Mathematics Supervisor: prof. Ing. Zdeněk Strakoš, DrSc. Abstract: After the derivation of the Conjugate Gradient method (CG) and the short review of its relationship with other fields of mathematics, this thesis is focused on its convergence behaviour both in exact and finite precision arith- metic. Fundamental difference between the CG and the Chebyshev semi-iterative method is described in detail. Then we investigate the use of the widespread lin- ear convergence bound based on Chebyshev polynomials. Through the example of the composite polynomial convergence bounds it is showed that the effects of rounding errors must be included in any consideration concerning the CG rate of convergence relevant to practical computations. Furthermore, the close corre- spondence between the trajectories of the CG approximations generated in finite precision and exact arithmetic is studied. The thesis is concluded with the discus- sion concerning the sensitivity of the closely related Gauss-Christoffel quadrature. The last two topics may motivate our further research. Keywords: Conjugate Gradient Method, Chebyshev semi-iterative method, fi- nite precision computations, delay of convergence, composite polynomial conver-...

Deep neural networks have achieved tremendous success in many domains (e.g., computer vision~\cite{Alexnet12,vggnet15,fastrcnn15}, speech recognition~\cite{hinton2012deep,dahl2012context}, natural language processing~\cite{dahl2012context,collobert2011natural}, games~\cite{silver2017mastering,silver2016mastering}), however, there are still many challenges in deep learning comunity such as how to speed up training large deep neural networks, how to compress large nerual networks for mobile/embed device without performance loss, how to automatically design the optimal network structures for a certain task, and how to further design the optimal networks with improved performance and certain model size with reduced computation cost. To speed up training large neural networks, we propose to use multinomial sampling for dropout, i.e., sampling features or neurons according to a multinomial distribution with different probabilities for different features/neurons. To exhibit the optimal dropout probabilities, we analyze the shallow learning with multinomial dropout and establish the risk bound for stochastic optimization. By minimizing a sampling dependent factor in the risk bound, we obtain a distribution-dependent dropout with sampling probabilities dependent on the second order statistics of the data distribution. To tackle the issue of evolving distribution of neurons in deep learning, we propose an efficient adaptive dropout (named evolutional dropout) that computes the sampling probabilities on-the-fly from a mini-batch of examples. To compress large neural network structures, we propose a simple yet powerful method for compressing the size of deep Convolutional Neural Networks (CNNs) based on parameter binarization. The striking difference from most previous work on parameter binarization/quantization lies at different treatments of $1\times 1$ convolutions and $k\times k$ convolutions ($k>1$), where we only binarize $k\times k$ convolutions into binary patterns. By doing this, we show that previous deep CNNs such as GoogLeNet and Inception-type Nets can be compressed dramatically with marginal drop in performance. Second, in light of the different functionalities of $1\times 1$ (data projection/transformation) and $k\times k$ convolutions (pattern extraction), we propose a new block structure codenamed the pattern residual block that adds transformed feature maps generated by $1\times 1$ convolutions to the pattern feature maps generated by $k\times k$ convolutions, based on which we design a small network with $\sim 1$ million parameters. Combining with our parameter binarization, we achieve better performance on ImageNet than using similar sized networks including recently released Google MobileNets. To automatically design neural networks, we study how to design a genetic programming approach for optimizing the structure of a CNN for a given task under limited computational resources yet without imposing strong restrictions on the search space. To reduce the computational costs, we propose two general strategies that are observed to be helpful: (i) aggressively selecting strongest individuals for survival and reproduction, and killing weaker individuals at a very early age; (ii) increasing mutation frequency to encourage diversity and faster evolution. The combined strategy with additional optimization techniques allows us to explore a large search space but with affordable computational costs. To further design the optimal networks with improved performance and certain model size under reduced computation cost, we propose an ecologically inspired genetic approach for neural network structure search , that includes two types of succession: primary and secondary succession as well as accelerated extinction. Specifically, we first use primary succession to rapidly evolve a community of poor initialized neural network structures into a more diverse community, followed by a secondary succession stage for fine-grained searching based on the networks from the primary succession. Accelerated extinction is applied in both stages to reduce computational cost. In addition, we also introduce the gene duplication to further utilize the novel block of layers that appeared in the discovered network structure.

O presente trabalho apresenta um estudo numérico dos efeitos da aplicação de geometrias geradoras de vórtices, com intuito de controlar passivamente a camada limite, em um perfil aerodinâmico que integra a asa traseira de multi elementos de um veículo de Fórmula SAE. As equações de Navier-Stokes com médias de Reynolds foram resolvidas utilizando o modelo k-ω SST (Shear Stress Transport) para o problema de fechamento da turbulência. Uma metodologia numérica padrão foi definida e utilizada nos diferentes casos analisados. Domínio de cálculo, malha, condições de contorno e critério de convergência foram escolhidos com base em norma SAE para análise numérica de escoamento externo em veículos terrestres. As camadas de volumes prismáticos próximos as superfícies com não-deslizamento foram dimensionadas de forma a resultar em um tratamento de parede adequado ao modelo de turbulência aplicado. O método GCI (Grid Convergence Index) foi utilizado para avaliar a qualidade da malha. Com o intuito de reduzir o custo computacional nos testes com diferentes configurações de geradores de vórtices, apenas parte de interesse do domínio de cálculo foi resolvido, impondo perfis de velocidade, energia cinética da turbulência e dissipação específica em sua entrada. Estas condições foram importadas da simulação com domínio completo resolvida Para verificar a correta captação dos principais efeitos físicos envolvidos, comparações com resultados experimentais foram feitas para 2 casos com escoamentos representativos: o corpo de Ahmed e um perfil aerodinâmico com geradores de vórtices. Além disso, as diferenças entre resolver o domínio completo ou parcial foram estudadas em outro comparativo com resultados experimentais. Concluiu-se que a metodologia numérica foi capaz de obter os coeficientes aerodinâmicos, e suas tendências frente a mudanças de geometria, nos casos estudados. Resolver parcialmente o domínio, impondo perfis em sua entrada, acarretou em diferença nos coeficientes obtidos na ordem de 2% para o coeficiente de sustentação e 7% para o coeficiente de arrasto. O controle passivo via geradores de vórtices foi eficaz em atrasar a separação da camada limite no flap do veículo de Fórmula SAE, as melhoras nos coeficientes de arrasto e sustentação foram da ordem de 7% e 0,3%, respectivamente. / The present work is a numerical study of the effects of the application of vortex generating geometries, in order to passively control the boundary layer, in an aerodynamic profile that integrates a multi-element rear wing of a Formula SAE vehicle. The Reynolds Averaged Navier-Stokes equations were solved using the k-ω Shear Stress Transport model for the turbulence closure problem. A standard numerical methodology was defined and used in the different cases analyzed. Computational domain, mesh, boundary conditions and convergence criteria were chosen based on SAE standard for numerical analysis of external flow in land vehicles. The layers of prismatic volumes near the non-slip surfaces were dimensioned to result in a wall treatment suitable to the applied turbulence model. The Grid Convergence Index (GCI) method was applied to evaluate the mesh quality. In order to reduce the computational cost in tests with different vortex generators configurations, only the part of interest of the calculation domain was solved, imposing velocity, turbulent kinetic energy and specific dissipation profiles on its inlet These conditions were imported from the full domain simulation already solved. To verify the correct capture of the main physical effects involved, comparisons with experimental results were made for 2 cases with representative flows: the Ahmed body and an aerodynamic profile with vortex generators. In addition, the differences between solving the complete or partial domain were studied in another comparative with experimental results. It was concluded that the numerical methodology was able to obtain the aerodynamic coefficients, and their tendencies against changes of geometry, in the cases studied. Partially solving the domain, imposing profiles at its entrance, resulted in a difference in the coefficients obtained in the order of 2% for the lift coefficient and 7% for the drag coefficient. The passive control via vortex generators was effective in delaying the separation of the boundary layer on the flap of the Formula SAE vehicle, the improvements in drag and lift coefficients were of the order of 7% and 0,3%, respectively.

O presente trabalho apresenta um estudo numérico dos efeitos da aplicação de geometrias geradoras de vórtices, com intuito de controlar passivamente a camada limite, em um perfil aerodinâmico que integra a asa traseira de multi elementos de um veículo de Fórmula SAE. As equações de Navier-Stokes com médias de Reynolds foram resolvidas utilizando o modelo k-ω SST (Shear Stress Transport) para o problema de fechamento da turbulência. Uma metodologia numérica padrão foi definida e utilizada nos diferentes casos analisados. Domínio de cálculo, malha, condições de contorno e critério de convergência foram escolhidos com base em norma SAE para análise numérica de escoamento externo em veículos terrestres. As camadas de volumes prismáticos próximos as superfícies com não-deslizamento foram dimensionadas de forma a resultar em um tratamento de parede adequado ao modelo de turbulência aplicado. O método GCI (Grid Convergence Index) foi utilizado para avaliar a qualidade da malha. Com o intuito de reduzir o custo computacional nos testes com diferentes configurações de geradores de vórtices, apenas parte de interesse do domínio de cálculo foi resolvido, impondo perfis de velocidade, energia cinética da turbulência e dissipação específica em sua entrada. Estas condições foram importadas da simulação com domínio completo resolvida Para verificar a correta captação dos principais efeitos físicos envolvidos, comparações com resultados experimentais foram feitas para 2 casos com escoamentos representativos: o corpo de Ahmed e um perfil aerodinâmico com geradores de vórtices. Além disso, as diferenças entre resolver o domínio completo ou parcial foram estudadas em outro comparativo com resultados experimentais. Concluiu-se que a metodologia numérica foi capaz de obter os coeficientes aerodinâmicos, e suas tendências frente a mudanças de geometria, nos casos estudados. Resolver parcialmente o domínio, impondo perfis em sua entrada, acarretou em diferença nos coeficientes obtidos na ordem de 2% para o coeficiente de sustentação e 7% para o coeficiente de arrasto. O controle passivo via geradores de vórtices foi eficaz em atrasar a separação da camada limite no flap do veículo de Fórmula SAE, as melhoras nos coeficientes de arrasto e sustentação foram da ordem de 7% e 0,3%, respectivamente. / The present work is a numerical study of the effects of the application of vortex generating geometries, in order to passively control the boundary layer, in an aerodynamic profile that integrates a multi-element rear wing of a Formula SAE vehicle. The Reynolds Averaged Navier-Stokes equations were solved using the k-ω Shear Stress Transport model for the turbulence closure problem. A standard numerical methodology was defined and used in the different cases analyzed. Computational domain, mesh, boundary conditions and convergence criteria were chosen based on SAE standard for numerical analysis of external flow in land vehicles. The layers of prismatic volumes near the non-slip surfaces were dimensioned to result in a wall treatment suitable to the applied turbulence model. The Grid Convergence Index (GCI) method was applied to evaluate the mesh quality. In order to reduce the computational cost in tests with different vortex generators configurations, only the part of interest of the calculation domain was solved, imposing velocity, turbulent kinetic energy and specific dissipation profiles on its inlet These conditions were imported from the full domain simulation already solved. To verify the correct capture of the main physical effects involved, comparisons with experimental results were made for 2 cases with representative flows: the Ahmed body and an aerodynamic profile with vortex generators. In addition, the differences between solving the complete or partial domain were studied in another comparative with experimental results. It was concluded that the numerical methodology was able to obtain the aerodynamic coefficients, and their tendencies against changes of geometry, in the cases studied. Partially solving the domain, imposing profiles at its entrance, resulted in a difference in the coefficients obtained in the order of 2% for the lift coefficient and 7% for the drag coefficient. The passive control via vortex generators was effective in delaying the separation of the boundary layer on the flap of the Formula SAE vehicle, the improvements in drag and lift coefficients were of the order of 7% and 0,3%, respectively.

O presente trabalho apresenta um estudo numérico dos efeitos da aplicação de geometrias geradoras de vórtices, com intuito de controlar passivamente a camada limite, em um perfil aerodinâmico que integra a asa traseira de multi elementos de um veículo de Fórmula SAE. As equações de Navier-Stokes com médias de Reynolds foram resolvidas utilizando o modelo k-ω SST (Shear Stress Transport) para o problema de fechamento da turbulência. Uma metodologia numérica padrão foi definida e utilizada nos diferentes casos analisados. Domínio de cálculo, malha, condições de contorno e critério de convergência foram escolhidos com base em norma SAE para análise numérica de escoamento externo em veículos terrestres. As camadas de volumes prismáticos próximos as superfícies com não-deslizamento foram dimensionadas de forma a resultar em um tratamento de parede adequado ao modelo de turbulência aplicado. O método GCI (Grid Convergence Index) foi utilizado para avaliar a qualidade da malha. Com o intuito de reduzir o custo computacional nos testes com diferentes configurações de geradores de vórtices, apenas parte de interesse do domínio de cálculo foi resolvido, impondo perfis de velocidade, energia cinética da turbulência e dissipação específica em sua entrada. Estas condições foram importadas da simulação com domínio completo resolvida Para verificar a correta captação dos principais efeitos físicos envolvidos, comparações com resultados experimentais foram feitas para 2 casos com escoamentos representativos: o corpo de Ahmed e um perfil aerodinâmico com geradores de vórtices. Além disso, as diferenças entre resolver o domínio completo ou parcial foram estudadas em outro comparativo com resultados experimentais. Concluiu-se que a metodologia numérica foi capaz de obter os coeficientes aerodinâmicos, e suas tendências frente a mudanças de geometria, nos casos estudados. Resolver parcialmente o domínio, impondo perfis em sua entrada, acarretou em diferença nos coeficientes obtidos na ordem de 2% para o coeficiente de sustentação e 7% para o coeficiente de arrasto. O controle passivo via geradores de vórtices foi eficaz em atrasar a separação da camada limite no flap do veículo de Fórmula SAE, as melhoras nos coeficientes de arrasto e sustentação foram da ordem de 7% e 0,3%, respectivamente. / The present work is a numerical study of the effects of the application of vortex generating geometries, in order to passively control the boundary layer, in an aerodynamic profile that integrates a multi-element rear wing of a Formula SAE vehicle. The Reynolds Averaged Navier-Stokes equations were solved using the k-ω Shear Stress Transport model for the turbulence closure problem. A standard numerical methodology was defined and used in the different cases analyzed. Computational domain, mesh, boundary conditions and convergence criteria were chosen based on SAE standard for numerical analysis of external flow in land vehicles. The layers of prismatic volumes near the non-slip surfaces were dimensioned to result in a wall treatment suitable to the applied turbulence model. The Grid Convergence Index (GCI) method was applied to evaluate the mesh quality. In order to reduce the computational cost in tests with different vortex generators configurations, only the part of interest of the calculation domain was solved, imposing velocity, turbulent kinetic energy and specific dissipation profiles on its inlet These conditions were imported from the full domain simulation already solved. To verify the correct capture of the main physical effects involved, comparisons with experimental results were made for 2 cases with representative flows: the Ahmed body and an aerodynamic profile with vortex generators. In addition, the differences between solving the complete or partial domain were studied in another comparative with experimental results. It was concluded that the numerical methodology was able to obtain the aerodynamic coefficients, and their tendencies against changes of geometry, in the cases studied. Partially solving the domain, imposing profiles at its entrance, resulted in a difference in the coefficients obtained in the order of 2% for the lift coefficient and 7% for the drag coefficient. The passive control via vortex generators was effective in delaying the separation of the boundary layer on the flap of the Formula SAE vehicle, the improvements in drag and lift coefficients were of the order of 7% and 0,3%, respectively.

<p>Uncertainty exists everywhere in scientific and engineering applications. To avoid potential risk, it is critical to understand the impact of uncertainty on a system by performing uncertainty quantification (UQ) and reliability analysis (RA). However, the computational cost may be unaffordable using current UQ methods with high-dimensional input. Moreover, current UQ methods are not applicable when numerical data and image data coexist. </p> <p>To decrease the computational cost to an affordable level and enable UQ with special high dimensional data (e.g. image), this dissertation develops three UQ methodologies with high dimensionality of input space. The first two methods focus on high-dimensional numerical input. The core strategy of Methodology 1 is fixing the unimportant variables at their first step most probable point (MPP) so that the dimensionality is reduced. An accurate RA method is used in the reduced space. The final reliability is obtained by accounting for the contributions of important and unimportant variables. Methodology 2 addresses the issue that the dimensionality cannot be reduced when most of the variables are important or when variables equally contribute to the system. Methodology 2 develops an efficient surrogate modeling method for high dimensional UQ using Generalized Sliced Inverse Regression (GSIR), Gaussian Process (GP)-based active learning, and importance sampling. A cost-efficient GP model is built in the latent space after dimension reduction by GSIR. And the failure boundary is identified through active learning that adds optimal training points iteratively. In Methodology 3, a Convolutional Neural Networks (CNN) based surrogate model (CNN-GP) is constructed for dealing with mixed numerical and image data. The numerical data are first converted into images and the converted images are then merged with existing image data. The merged images are fed to CNN for training. Then, we use the latent variables of the CNN model to integrate CNN with GP to quantify the model error using epistemic uncertainty. Both epistemic uncertainty and aleatory uncertainty are considered in uncertainty propagation. </p> <p>The simulation results indicate that the first two methodologies can not only improve the efficiency but also maintain adequate accuracy for the problems with high-dimensional numerical input. GSIR with active learning can handle the situations that the dimensionality cannot be reduced when most of the variables are important or the importance of variables are close. The two methodologies can be combined as a two-stage dimension reduction for high-dimensional numerical input. The third method, CNN-GP, is capable of dealing with special high-dimensional input, mixed numerical and image data, with the satisfying regression accuracy and providing an estimate of the model error. Uncertainty propagation considering both epistemic uncertainty and aleatory uncertainty provides better accuracy. The proposed methods could be potentially applied to engineering design and decision making. </p>

Kubernetes is a free, open-source container orchestration system for deploying and managing Docker containers that host microservices. Its cluster logs are extremely helpful in determining the root cause of a failure. However, as systems become more complex, locating failures becomes more difficult and time-consuming. This study aims to identify the classification algorithms that accurately classify the given log data and, at the same time, require fewer computational resources. Because the data is quite large, we begin with expert-based feature selection to reduce the data size. Following that, TF-IDF feature extraction is performed, and finally, we compare five classification algorithms, SVM, KNN, random forest, gradient boosting and MLP using several metrics. The results show that Random forest produces good accuracy while requiring fewer computational resources compared to other algorithms.

Context: The goal of this research is to form a correlation between code packages and test cases which is done by using automated weak mutation. The correlations formed is used as the statistical test data for selecting relevant tests from the test suite which decreases the size of the test suite and speed up the process. Objectives: In this study, we have done an investigation of existing methods for reducing the computational cost of automatic mutation testing. After the investigation, we build an open source automatic mutation tool that mutates the source code to run on the test cases of the mutated code that maps the failed test to the part of the code that was changed. The failed test cases give the correlation between the test and the source code which is collected as data for future use of the test selection. Methods: Literature review and Experimentation is chosen for this research. It was a controlled experiment done at the Swedish ICT company to mutate the camera codes and test them using the regression test suite. The camera codes provided are from the continuous integration of historical data. We have chosen experimentation as our research because as this method of research is more focused on analyzing the data and implementing a tool using historical data. A literature review is done to know what kind of mutation testing reduces the computational cost of the testing process. The implementation of this process is done by using experimentation Results: The comparative results obtained after mutating the source code with regular mutants and weak mutants we have found that regular mutants and weak mutants are compared with their correlation accuracy and we found that on regular mutation operators we got 62.1% correlation accuracy and coming to weak mutation operators we got 85% of the correlation accuracy. Conclusions: This research on experimentation to form the correlations in generating test selection statistics using automated mutation testing in the continuous integration environment for improving test cases selection in regression testing

The laterally-loaded pile has long been a topic of research interest. Several models of the soil surrounding a pile have been developed for simulation of lateral pile behavior, ranging from simple spring and dashpot models to sophisticated three-dimensional finite-element models. However, results from the available pile-soil models are not accurate due to inherent approximations or constraints. For the springs and dashpots representation, the real and imaginary stiffness are calculated by idealizing the soil domain as a series of plane-strain slices leading to unrealistic pile behavior at low frequencies while the three-dimensional finite-element analysis is very computationally demanding. Therefore, this dissertation research seeks to contribute toward procedures that are computationally cost-effective while accuracy of the computed response is maintained identical or close to that of the three-dimensional finite-element solution. Based on the fact that purely-elastic soil displacement variations in azimuthal direction are known, the surrounding soil can be formulated in terms of an equivalent one-dimensional model leading to a significant reduction of computational cost. The pile with conventional soil-slice model will be explored first. Next, models with shear stresses between soil slices, including and neglecting the soil vertical displacement, are investigated. Excellent agreement of results from the proposed models with three-dimensional finite-element solutions can be achieved with only small additional computational cost. / text