Computed tomography (CT) combined with deep learning (DL) has recently shown great potential in biomedical imaging. Complex DL models with varying architectures inspired by the human brain are improving imaging software and aiding diagnosis. However, the accuracy of these DL models heavily relies on the datasets used for training, which often contain low-quality CT images from low-dose CT (LDCT) scans. Moreover, in contrast to the neural architecture of the human brain, DL models today are dense and complex, resulting in a significant computational footprint. Therefore, in this work, we propose sparse optimizations to minimize the complexity of the DL models and leverage architecture-aware optimization to reduce the total training time of these DL models. To that end, we leverage a DL model called DenseNet and Deconvolution Network (DDNet). The model enhances LDCT chest images into high-quality (HQ) ones but requires many hours to train. To further improve the quality of final HQ images, we first modified DDNet's architecture with a more robust multi-level VGG (ML-VGG) loss function to achieve state-of-the-art CT image enhancement. However, improving the loss function results in increased computational cost. Hence, we introduce sparse optimizations to reduce the complexity of the improved DL model and then propose architecture-aware optimizations to efficiently utilize the underlying computing hardware to reduce the overall training time. Finally, we evaluate our techniques for performance and accuracy using state-of-the-art hardware resources. / Master of Science / Deep learning-based (DL) techniques that leverage computed tomography (CT) are becoming omnipresent in diagnosing diseases and abnormalities associated with different parts of the human body. However, their diagnostic accuracy is directly proportional to the quality of the CT images used in training the DL models, which is majorly governed by the radiation dose of the X-ray in the CT scanner. To improve the quality of low-dose CT (LDCT) images, DL-based techniques show promising improvements. However, these techniques require substantial computational resources and time to train the DL models. Therefore, in this work, we incorporate algorithmic techniques inspired by sparse neural architecture of the human brain to reduce the complexity of such DL models. To that end, we leverage a DL model called DenseNet and Deconvolution Network (DDNet) that enhances the quality of CT images generated by low X-ray dosage into high-quality CT images. However, due to its architecture, it takes hours to train DDNet on state-of-the-art hardware resources. Hence, in this work, we propose techniques that efficiently utilize the hardware resources and reduce the time required to train DDNet. We evaluate the efficacy of our techniques on modern supercomputers in terms of speed and accuracy.

This thesis delves into the advancement of 5G portable test-nodes in wireless communication systems with cognitive radio capabilities, specifically addressing the critical need for dynamic spectrum sensing and awareness at the radio receiver through AI-driven automatic modulation classification. Our methodology is centered around the transformer encoder architecture incorporating a multi-head self-attention mechanism. We train our architecture extensively across a diverse range of signal-to-noise ratios (SNRs) from the RadioML 2018.01A dataset. We introduce a novel transformer-based architecture with a gated mechanism, designed as a runtime re-configurable automatic modulation classification framework, which demonstrates enhanced performance with low SNR RF signals during evaluation, an area where conventional methods have shown limitations, as corroborated by existing research. Our innovative single-model framework employs distinct weight sets, activated by varying SNR levels, to enable a gating mechanism for more accurate modulation classification. This advancement in automatic modulation classification marks a crucial step toward the evolution of smarter communication systems. / Master of Science / This thesis delves into the advancement of wireless communication systems, particularly in developing portable devices capable of effectively detecting and analyzing radio signals with cognitive radio capabilities. Central to our research is leveraging artificial intelligence (AI) for automatic modulation classification, a method to identify signal modulation types. We utilize a transformer-based AI model trained on the RadioML 2018.01A dataset. Our training approach is particularly effective when evaluating low-quality signals using a gating mechanism based on signal-to-noise ratios, an area previously considered challenging in existing research. This work marks a significant advancement in creating more intelligent and responsive wireless communication systems.

As the field of mobile robotics rapidly expands, precise understanding of a robot’s position and orientation becomes critical for autonomous navigation and efficient task performance. In this thesis, we present a snapshot-based global localization machine learning model for a mobile robot, the e-puck, in a simulated environment. Our model uses multimodal data to predict both position and orientation using the robot’s on-board cameras and LiDAR sensor. In an effort to minimize localization error, we explore different sensor configurations by varying the number of cameras and LiDAR layers used. Additionally, we investigate the performance benefits of different multimodal fusion strategies while leveraging the EfficientNet CNN architecture as our model’s foundation. Data collection and testing is conducted using Webots simulation software, and our results show that, when tested in a 12m x 12m simulated apartment environment, our model is able to achieve positional accuracy within 0.2m for each of the x and y coordinates and orientation accuracy within 2°, all without the need for sequential data history. Our results demonstrate the potential for accurate global localization of mobile robots in simulated environments without the need for existing maps or temporal data.

In this work, we examine semi-supervised learning for Gait recognition with a limited number of labeled samples. Our research focus on two distinct aspects for limited labels, 1)closed-set: with limited labeled samples per individual, and 2) open-set: with limited labeled individuals. We find open-set poses greater challenge compared to closed-set thus, having more labeled ids is important for performance than having more labeled samples per id. Moreover, obtaining labeled samples for a large number of individuals is usually more challenging, therefore limited id setup (closed-setup) is more important to study where most of the training samples belong to unknown ids. We further analyze that existing semi-supervised learning approaches are not well suited for scenario where unlabeled samples belong to novel ids. We propose a simple prototypical self-training approach to solve this problem, where, we integrate semi-supervised learning for closed set setting with self-training which can effectively utilize unlabeled samples from unknown ids. To further alleviate the challenges of limited labeled samples, we explore the role of synthetic data where we utilize diffusion model to generate samples from both known and unknown ids. We perform our experiments on two different Gait recognition benchmarks, CASIA-B and OUMVLP, and provide a comprehensive evaluation of the proposed method. The proposed approach is effective and generalizable for both closed and open-set settings. With merely 20% of labeled samples, we were able to achieve performance competitive to supervised methods utilizing 100% labeled samples while outperforming existing semi-supervised methods.

In this thesis we aim to supplement the Climate and Economic Justice Screening Tool (CE JST), which assists federal agencies in identifying disadvantaged census tracts, by extracting five environmental features from Google Street View (GSV) images. The five environmental features are garbage bags, greenery, and three distinct road damage types (longitudinal, transverse, and alligator cracks), which were identified using image classification, object detection, and image segmentation. We evaluate three cities using this developed feature space in order to distinguish between disadvantaged and non-disadvantaged census tracts. The results of the analysis reveal the significance of the feature space and demonstrate the time efficiency, detail, and cost-effectiveness of the proposed methodology. / Master of Science / In this thesis we aim to supplement the Climate and Economic Justice Screening Tool (CE JST), which assists federal agencies in identifying disadvantaged census tracts, by extracting five environmental features from Google Street View (GSV) images. The five environmental features are garbage bags, greenery, and three distinct road damage types (longitudinal, transverse, and alligator cracks), which were identified using image classification, object detection, and image segmentation. We evaluate three cities using this developed feature space in order to distinguish between disadvantaged and non-disadvantaged census tracts. The results of the analysis reveal the significance of the feature space and demonstrate the time efficiency, detail, and cost-effectiveness of the proposed methodology.

This paper discusses how Q-Learning and Deep Q-Networks (DQN) canbe applied to state-action problems described by a Markov decision process(MDP). These are machine learning methods for finding the optimal choiceof action at each time step, resulting in the optimal policy. The limitationsand advantages for the two methods are discussed, with the main limitationbeing the fact that Q-learning is unable to be used on problems with infinitestate spaces. Q-learning, however, has an advantage in the simplicity of thealgorithm, leading to a better understanding of what the algorithm is actuallydoing. Q-Learning did manage to find the optimal policy for the simpleproblem studied in this paper, but was unable to do so for the advancedproblem. The Deep Q-Network (DQN) approach was able to solve bothproblems, with a drawback in it being harder to understand what the algorithmactually is doing.

This thesis investigates the performance of deep learning models, specifically Resnet50 and TransUnet, in semantic image segmentation on microwell images containing tumor and natural killer (NK) cells. The main goal is to examine the effect of only using bright-field data (1-channel) as input instead of both fluorescent and brightfield data (4-channel); this is interesting since fluorescent imaging can cause damage to the cells being analyzed. The network performance is measured by Intersection over Union (IoU), the networks were trained and using manually annotated data from Onfelt Lab. TransUnet consistently outperformed the Resnet50 for both the 4-channel and 1-channel data. Moreover, the 4-channel input generally resulted in a better IoU compared to using only the bright-field channel. Furthermore, a significant decline in performance is observed when the networks are tested on the control data. For the control data, the overall IoU for the best performing 4-channel model dropped from 86.2\% to 73.9\%. The best performing 1-channel model dropped from 83.8\% to 70.8\% overall IoU.

The Advanced Scatterometer (ASCAT) is a satellite-based remote sensing instrument designed for measuring wind speed and direction over the Earth's oceans. This thesis aims to expand and improve the capabilities of ASCAT by adding rain detection and advancing wind retrieval. Additionally, this expansion to ASCAT serves as evidence of Artificial Intelligence (AI) techniques learning both novel and traditional methods in remote sensing. I apply semantic segmentation to ASCAT measurements to detect rain over the oceans, enhancing capabilities to monitor global precipitation. I use two common neural network architectures and train them on measurements from the Tropical Rainfall Measuring Mission (TRMM) collocated with ASCAT measurements. I apply the same semantic segmentation techniques on wind retrieval in order to create a machine learning model that acts as an inverse Geophysical Model Function (GMF). I use three common neural network architectures and train the models on ASCAT data collocated with European Centre for Medium-Range Weather Forecasts (ECMWF) wind vector data. I successfully increase the capabilities of the ASCAT satellite to detect rainfall in Earth's oceans, with the ability to retrieve wind vectors without a GMF or Maximum Likelihood Estimation (MLE).

Einleitung: Ziel der Arbeit war es, die tiefen Halsinfektionen im Hals-, Nasen- und Ohrenbereich anhand der Ätiologie, diagnostischer Verfahren, Klinik und Therapie mit der aktuellen internationalen Literatur zu vergleichen. Material&Methode: Hierzu erfolgte eine retrospektive Auswertung von 63 Patienten mit tiefen Halsabszessen, die im Zeitraum zwischen Januar 2002 und Dezember 2012 an der Universitätsmedizin Göttingen in der Klinik für HNO-Heilkunde behandelt wurden. Die statistische Asuwertung erfolgte dekriptiv sowie analysierend mit graphischen Darstellungsformen. Die metrischen Variablen wurde mittels Mann-Whitney-U-Test sowie dem Exakte Fisher-Test auf Signifikanz (p=0,05) getestet. Ergebnisse: Tiefe Halsabszesse sind am häufigsten im Spatium parapharygeum anzutreffen. Streptococcus viridans (26,7%), meist als Mischinfektion mit anaeroben Bakterien, ist der häufigste Erreger tiefer Halsinfektionen. Das Keimspektrum unterscheidet sich signifikant beim Krankheitsbild "Diabetes mellitus", bei dem Staphylococcus aureus als häufigster Keim identifiziert wurde (p=0,02). Zusammenfassung: Die Therapie der Wahl bei abszedierenden tiefen Halsinfektionen bleibt die frühzeitige chirurgische Sanierung mit Abszesseröffnung und Drainage sowie die Sicherung der Atemwege in Verbindung mit einer gezielten intravenösen Antibiotikatherapie. Außerdem sollte bei jeder tiefen Halsinfektion eine standardisierte Erreger-bestimmung - inklusive Anitibiogramm- gefordert werden.

O presente trabalho de mestrado está inserido nos estudos de refração profunda do Projeto Temático 'Estudos Geofísicos e Modelo Tectônico dos Setores Central e Sudeste da Província Tocantins, Brasil Central'. Nesses estudos foram levantadas três linhas de refração de aproximadamente 300 km de extensão, duas no setor Central da Província Tocantins e uma no setor Sudeste, que é o objeto de estudo deste trabalho. Foram utilizados 111 sismógrafos digitais SGR pertencentes ao programa PASSCAL, instrumentos auxiliares do USGS, e 13 sismógrafos digitais e instrumentos auxiliares do IAG/USP. A linha sísmica teve aproximadamente 300 km de extensão com pontos de registro separados a cada 2,5 km, distribuídos ao longo de estradas principais e secundárias. A cada 50 km, aproximadamente, foi realizada uma explosão, nas explosões dos extremos da linha foram utilizados 1000 kg de explosivo e para a explosão central uma carga de 500 kg. Para a determinação das coordenadas geográficas dos pontos de tiro e de registro, foi utilizado o método diferencial com medidas de GPS. O principal objetivo deste trabalho foi obter como produto final um modelo de velocidades sísmicas contendo as características físicas das principais descontinuidades na crosta terrestre e no manto superior. Para análise e processamento dos dados foram utilizados os pacotes SAC, SU, SEIS. Para a modelagem foram utilizados a teoria do raio e a elaboração de sismogramas sintéticos, do pacote SEIS. Para a elaboração do modelo final foram utilizados os dados das explosões dos pontos extremos e central, tendo em vista que devido a problemas técnicos não foram registrados os sinais das outras 4 explosões. Além disso, as explosões registradas não apresentaram sinais claros em toda a extensão da linha. Devido a tudo isso e considerando as unidades geológicas presentes na região de estudo são sugeridos três modelos de velocidades sísmicas. O primeiro modelo refere-se ao tiro direto (EX31) localizado no extremo sudoeste da linha, sobre a Bacia do Paraná. Para este modelo obteve-se para superfície (0 km) a velocidade inicial de 2 km/s (coberturas); para a profundidade de 0,086 km a velocidade inicial é de 5,15 km/s (basalto); para a profundidade de 0,350 km obteve-se a velocidade inicial de 4,6 km/s (arenito - camada de baixa velocidade); para profundidade de 0,650 km a velocidade inicial é de 5,75 km/s e para profundidade de 4 km obteve-se a velocidade inicial de 6,07 km/s. O segundo modelo refere-se ao tiro reverso (EX34) localizado no centro da linha sobre granitóides do Grupo Araxá. Para este modelo obteve -se para superfície (0 km) a velocidade inicial de 2 km/s; para a profundidade de 0,06 km a velocidade inicial é de 5,69 km/s e para a profundidade de 0,860 km obteve-se a velocidade inicial de 6,25 km/s. Finalmente, o terceiro modelo refere-se ao tiro direto para toda a extensão da linha (300 km). Este modelo foi definido a partir de fases secundárias lidas nos registros e modelos anteriores propostos na literatura. Da superfície até os 4 km iniciais de profundidade este modelo é igual ao primeiro, para uma profundidade de 20 km obteve-se a velocidade inicial de 6,70 km/s e para uma profundidade de 40 km a velocidade é de 8,00 km/s (descontinuidade de MOHO). / This work to fulfil the degree of Master of Sciences is inserted among the deep seismic refraction studies of the Thematic Project 'Geophysical Studies and Tectonic Model of the Tocantins Province Central and Southeast Sectors, Central Brazil'. Three refraction lines, of around 300 km long each, were deployed, two of them in the Central sector and the other in the SE sector, that is subject of the present work. The equipment used in this experiment was composed by 111 SGR digital seismographs belonging to the PASSCAL Program. Complemented with auxiliary instruments from USGS and 13 seismographs belonging to IAG/USP. The space among the recording points was 2.5 km, which were located along main and secondary roads. Every 50 km was fired an explosion with 1000 kg of emulsion in each extreme and 500 kg in the central point. The geographical co-ordinates were determined by using the GPS differential method. The main objective of this work is to obtain as a final product a seismic velocity model with the physical characteristics of the main discontinuities in the crust and upper mantle. The packages SAC, SU and SEIS were used to perform the data analysis and processing. To carry on the modelling were used the ray theory and the synthetic seismograms construction, belonging to the SEIS package Data from the extreme and middle points of the seismic line were used to elaborate the final model, considering that due to technical problems signals from the other four explosions were not recorded. Apart from that, the recorded explosions did not present clear signals all along the extension of the line. Due to these facts, and considering also the geological units present in the studied region, are suggested three seismic velocity models. The first model is referred to the direct shot (EX31), which is localised in the Southwest extreme of the line on the Parana Basin province. In this model we obtained the P wave velocity (VP) of 2 km/sec at the surface, corresponding to the unconsolidated sediments and soil on the top of that basin. At a depth of 86 m we found VP of 5,15 km/sec and at a depth of 350 m the velocity VP of 4,6 km/sec, corresponding to the basalt and sand layers of the Parana Basin. Underlying them, at 650 m of depth we found the basement with VP of 5,75 km/sec and finally at a depth of 4 km there is a layer with VP of 6,07 km/sec, corresponding to a typical upper crust P wave velocity. The second model corresponds to the reverse shot (EX34) that is localised in the middle point of the line on the granitoides of the Araxa Group. For this model we obtained VP of 2 km/sec for the superficial layers, then at a depth of 60 m was obtained V P of 5,69 km/sec and for a depth of 860 m the value of V P is 6,25 km/sec. Finally, the third model belongs to the whole line section (300 km) from the direct shot (EX31). This model was obtained by using the arrivals of secondary phases and the results of models proposed in other works. From the surface down to 4 km of depth this model is similar to the first one. At 20 km of depth there is a layer with VP of 6,70 km/sec, corresponding to the lower crust, with Moho at a depth of 40 km with VP of 8,00 km/sec.