Diese Dissertation befasst sich mit räumlichen Wahrnehmungen und Diskursen, mit denen man den Raum und seine Bestandteile behandelte. Die Eroberung Sibiriens im 17. Jahrhundert bewirkte einen tiefgreifenden Wandel in den russischen Vorstellungen über die weit entfernte Peripherie sowie deren Ressourcen. Die neuen Denkweisen kristallisierten sich in einer diskursiven Formation heraus, die Macht über Raum und Rohstoffe Sibiriens symbolisierte und organisierte. Dieser „Berg-Diskurs“ trug moderne Züge, denn er bedurfte sich neuer Formen der Kontrolle über die Raumsproduktion. Diese Einstellung wurde allmählich zu einer erstaunlich überlebensfähigen räumlichen Ideologie und zum festen Bestandteil des russischen Bodenschätzediskurses der Zukunft. Die Rolle der Wissensproduzenten wechselte zwischen den zentralen und regionalen Institutionen und Netzwerken. Der „Altai“, der den kaiserlichen Bergbau-Bezirk und die Gebirgslandschaft umfasste, wurde auf Grund seines Rohstoffreichtums von Repräsentanten des russischen Staates als Region erfunden. Die Dissertation stellt die imaginären und realen Geographien des Altai in drei unterschiedlichen Dimensionen dar. Dabei geht es um den Wandel der Repräsentationen von geographischen Räumen und der Berglandschaften in Russland insgesamt (Makroebene), die Mehrschichtigkeit des russischen Diskurses über Bergregionen und Gebirgslandschaften (Mesoebene) und den Altai als facettenreiches Konzept einer komplexen imperialen geographischen Imagination (Mikroebene). Die Beschreibung des Altai faßte in sich zahlreiche inkohärente Bilder verschiedener sozialer Gruppen. Der Ort wurde durch mentale Geographien erfolgreich instrumentalisiert, z.B. „die Goldenen Gebirge“ und „die sibirische Schweiz“. Diese Bilder machten die Region sichtbar, sowohl für nationalistisch gesinnte Gruppen als auch die breiteren Bervölkerungsschichten. / This dissertation focuses on the production of imperial space with a particular emphasis on the role of power discourses concerning mineral resources. By relying on published materials, it aims to establish a new conceptual framework for the examining of cultural patterns and practices of imagining of space and mineral wealth. For that purpose, it introduces a concept of the ”Berg-Discourse” that expands our understanding of the Russian engagement with geographical space. It begins by exploring Russian exposure to the mountains and mineral resources of Siberia in terms of the spatial knowledge production. It then examines how Russian imperial strategies and aspirations were embedded in the making of the Altai, a vast mining territory in West Siberia that once formed a private domain of the Russian rulers. The dissertation argues that the making of the Altai was in many ways part of the same imperial impulse towards mineral exploitation. It explores the ways in which the Altai was imagined through its enormous mineral endowment; how the imagined place became real; and how this real place became imagined from various vantage points. As the study shows, the region acquired multiple mental representations, enjoying a near mythological presence across imperial culture. Finally, the dissertation concludes by showing how this landscape was incorporated into imperial and national myths in the course of production and consumption of spatial knowledge about the remote location.

Indiana University-Purdue University Indianapolis (IUPUI) / Research work presented within this thesis propose a neoteric variant of deep convolutional neural network architecture, CondenseNeXt, designed specifically for ARM-based embedded computing platforms with constrained computational resources. CondenseNeXt is an improved version of CondenseNet, the baseline architecture whose roots can be traced back to ResNet. CondeseNeXt replaces group convolutions in CondenseNet with depthwise separable convolutions and introduces group-wise pruning, a model compression technique, to prune (remove) redundant and insignificant elements that either are irrelevant or do not affect performance of the network upon disposition. Cardinality, a new dimension to the existing spatial dimensions, and class-balanced focal loss function, a weighting factor inversely proportional to the number of samples, has been incorporated in order to relieve the harsh effects of pruning, into the design of CondenseNeXt’s algorithm. Furthermore, extensive analyses of this novel CNN architecture was performed on three benchmarking image datasets: CIFAR-10, CIFAR-100 and ImageNet by deploying the trained weight on to an ARM-based embedded computing platform: NXP BlueBox 2.0, for real-time image classification. The outputs are observed in real-time in RTMaps Remote Studio’s console to verify the correctness of classes being predicted. CondenseNeXt achieves state-of-the-art image classification performance on three benchmark datasets including CIFAR-10 (4.79% top-1 error), CIFAR-100 (21.98% top-1 error) and ImageNet (7.91% single model, single crop top-5 error), and up to 59.98% reduction in forward FLOPs compared to CondenseNet. CondenseNeXt can also achieve a final trained model size of 2.9 MB, however at the cost of 2.26% in accuracy loss. Thus, performing image classification on ARM-Based computing platforms without requiring a CUDA enabled GPU support, with outstanding efficiency.

The diploma thesis deals with machine learning, expert systems, fuzzy logic, genetic algorithms, neural networks and chaos theory, which fall into the category of artificial intelligence. The aim of this work is to describe and implement three different classification methods, according to which the data set will be processed. The GNU Octave software environment was chosen for the data application for licensing reasons. Further evaluate the success of data classification, including visualization. Three different classification methods are used for comparison, so that we can compare the processed data with each other.

This thesis deals with classification of different types of time courses of ECG signals. Main objective was to recognize the normal cycles and several forms of arrhythmia and to classify the exact types of them. Classification has been done with usage of algorithms of Neural Networks in Matlab program, with its add-on (Neural Network Toolbox). The result of this thesis is application, which makes possible to load an ECG signal, pre-process it and classify its each cycle into five classes. Percentage results of this classification are in the conclusion of this thesis.

This master's thesis deals with data mining techniques and classification of the data into specified categories. The goal of this thesis is to implement a web portal for administration and classification of data from distributed sources. To achieve the goal, it is necessary to test different methods and find the most appropriate one for web articles classification. From the results obtained, there will be developed an automated application for downloading and classification of data from different sources, which would ultimately be able to substitute a user, who would process all the tasks manually.

The thesis deals with analysis and design of system for automatic localization and recognition of the license plate. The input images are from different sources, and contain large scenic and weather variations. The aim was to create a system able to find the licence plate on the image and recognize its alphanumeric figure. In this work, there is a focus on analysis and implementation of localization and optical character recognition methods. One own and four other localization methods are compared. There are also compared three classifiers for optical character recognition. Localization and OCR methods are tested on real data and evaluated in accordance with the calculated evaluation parameters. The work also contains sensitivity analysis of the proposed system.

The thesis is focused on analysis of polysomnographic signals based on extraction of chosen parameters in time, frequency and time-frequency domain. The parameters are acquired from 30 seconds long segments of EEG, EMG and EOG signals recorded during different sleep stages. The parameters used for automatic classification of sleep stages are selected according to statistical analysis. The classification is realized by artificial neural networks, k-NN classifier and linear discriminant analysis. The program with a graphical user interface was created using Matlab.

Neutrino astronomy is expanding into the ultra-high energy (>1017eV) frontier with the use of in-ice detection of Askaryan radio emission from neutrino-induced particle showers. There are already pilot arrays for validating the technology and the next few years will see the planning and construction of IceCube-Gen2, an upgrade to the current neutrino telescope IceCube. This thesis aims to facilitate that planning by providing faster simulations using deep learning surrogate models. Faster simulations could enable proper optimisation of the antenna stations providing better sensitivity and reconstruction of neutrino properties. The surrogates are made for two parts of the end-to-end simulations: the signal generation and the signal propagation. These two steps are the most time-consuming parts of the simulations. The signal propagation is modelled with a standard fully connected neural network whereas for the signal generation a conditional Wasserstein generative adversarial network is used. There are multiple reasons for using these types of models. For both problems the neural networks provide the speed necessary as well as being differentiable -both important factors for optimisation. Generative adversarial networks are used in the signal generation because of the inherent stochasticity in the particle shower development that leads to the Askaryan radio signal. A more standard neural network is used for the signal propagation as it is a regression task. Promising results are obtained for both tasks. The signal propagation surrogate model can predict the parameters of interest at the desired accuracy, except for the travel time which needs further optimisation to reduce the uncertainty from 0.5 ns to 0.1 ns. The signal generation surrogate model predicts the Askaryan emission well for the limited parameter space of hadronic showers and within 5° of the Cherenkov cone. The two models provide a first step and a proof of concept. It is believed that the models can reach the required accuracies with more work.

The development of machine learning research has provided statistical innovations and further developments within the field of time series analysis. This study seeks to investigate two different approaches on artificial neural network models based on different learning techniques, and answering how well the neural network approach compares with a basic autoregressive approach, as well as how the artificial neural network models compare to each other. The models were compared and analyzed in regards to the univariate forecast accuracy on 20 randomly drawn time series from two different time frequencies from the M4 competition dataset. Forecasting was made dependent on one time lag (t-1) and forecasted three and six steps ahead respectively. The artificial neural network models outperformed the baseline Autoregressive model, showing notably lower mean average percentage error overall. The Multilayered perceptron models performed better than the Long short-term memory model overall, whereas the Long short-term memory model showed improvement on longer prediction time dimensions. As the training were done univariately  on a limited set of time steps, it is believed that the one layered-approach gave a good enough approximation on the data, whereas the added layer couldn’t fully utilize its strengths of processing power. Likewise, the Long short-term memory model couldn’t fully demonstrate the advantagements of recurrent learning. Using the same dataset, further studies could be made with another approach to data processing. Implementing an unsupervised approach of clustering the data before analysis, the same models could be tested with multivariate analysis on models trained on multiple time series simultaneously.

The main goal of this master's thesis was to study the main principles of classification methods. Basic principles of knowledge discovery process, data mining and using an external class CSSBox are described. Special attantion was paid to implementation of a ,,k-nearest neighbors`` classification method. The first objective of this work was to create training and testing data described by 'n' attributes. The second objective was to perform experimental analysis to determine a good value for 'k', the number of neighbors.