Das Monitoring von Landwirtschaftsflächen ist eines der Kernthemen der zukünftigen EnMAP Mission, einem deutschen, hyperspektralen Fernerkundungssensor, dessen Start für 2015 geplant ist. In Vorbereitung dieser Mission gehören die Erweiterung und Entwicklung objektiver, robuster sowie zuverlässiger Methoden zur Ableitung biophysikalischer Parameter zu den Hauptaufgaben. Die für das Monitoring von Vegetation notwendige hohe zeitliche Auflösung wird durch ein stufenloses Schwenken von bis zu +/-30° quer zur Flugrichtung erreicht. Daraus resultiert, dass die Daten durch variierende Ein- und Ausstrahlungsgeometrien stark beeinflusst werden. Daher ist eine detaillierte Kenntnis der bidirektionalen Reflexionsfunktion (engl. bidirectional reflectance distribution function, BRDF) der beobachteten Oberflächen notwendig, um diese Einflüsse auf das Reflexionssignal zu identifizieren und anschließend zu korrigieren. Zu diesem Zweck wurde in dieser Arbeit eine Methodik entwickelt, die es ermöglicht, auf der Basis von simulierten Spektren realistische BRDF Szenarien zu modellieren und oberflächenspezifische Korrekturfunktionen abzuleiten. Die Methodik besteht aus drei aufeinander aufbauenden Komponenten. Im ersten Schritt erfolgt die Modellierung der BRDF von landwirtschaftlichen Vegetationsbeständen. Im zweiten Schritt wird der Einfluss der Bestandsarchitektur auf die BRDF analysiert. Darüber hinaus wird untersucht, inwiefern sich Variationen in der BRDF auf die quantitative Ausprägung von Vegetationsindizes auswirken. Solche Indizes sind eine häufig genutzte Möglichkeit zur Quantifizierung biophysikalische Parameter im Rahmen empirischer Verfahren. Aufbauend auf den gewonnenen Erkenntnissen wurden im dritten Schritt Korrekturfunktionen für ausgewählte Vegetationsindizes entwickelt, um Schrägblickbeobachtungen in Nadirbeobachtungen zu transformieren. Abschließend wurde die entwickelte Methodik auf simulierte, sensorspezifische Spektren übertragen. / Monitoring of arable crops is one of the core applications of the upcoming spaceborne EnMAP mission, a German hyperspectral imaging spectrometer scheduled for launch in 2015. During the present preparatory phase one of the primary tasks is the development of accurate, robust and reliable retrieval methods for biophysical canopy parameters. Monitoring of crop canopies requires a frequent temporal coverage. In case of EnMAP, this will be realised by an off-nadir pointing of the sensor up to +/-30° across to the flight direction. The off-nadir pointing leads to data strongly influenced by varying acquisition geometry. Therefore, detailed knowledge of bidirectional reflectance distribution functions (BRDF) of the observed surfaces is necessary to identify and to correct BRDF influenced reflectance signals. For this purpose, a methodology was developed that allows modelling of realistic BRDF scenarios and of surface-specific correction functions. This methodology consists of three consecutive parts. In the first part, modelling of the BRDF of crop canopies was performed. In the second part, the influence of canopy architecture on the BRDF was analysed. Additionally, the BRDF related dependencies of vegetation indices were investigated. Such indices are widely used to quantify biophysical canopy parameters based on empirical methods. In the third part, correction functions were developed for selected vegetation indices to enable a transformation from off-nadir into nadir observations. These correction functions incorporate results obtained in the second part of the methodology. Finally, the developed methodology was applied on sensor-specific simulated spectra.

This study focuses on the use of machine learning (ML) techniques to automate the detection of Cortical Evoked Auditory Responses (CEARs), which are key in understanding how the auditory cortex processes sound stimuli. Traditionally, analyzing these auditory responses has relied on manual interpretation by audiologists, a process that can introduce variability and human error, particularly in complex cases. To address this challenge, the research utilizes advanced deep learning models, including Convolutional Neural Networks (CNNs), Long Short Term Memory (LSTM) networks, and Bidirectional LSTM (BiLSTM) architectures, to analyze Electroencephalography (EEG) data and classify the presence or absence of auditory responses automatically. By employing these models, the study demonstrates improved accuracy in detecting auditory responses, with the BiLSTM model achieving the highest accuracy of 90%. Additionally, the use of Grad-CAM visualizations enables better interpretability of the model's predictions, allowing for insights into the biological relevance of the EEG features the models focused on. The findings highlight the potential of ML techniques to enhance the efficiency and accuracy of auditory diagnostics, which can support audiologists in clinical decision-making. The research also paves the way for future developments, such as integrating these models into real-time EEG systems and expanding their use to other time-series data or domains like speech recognition or ECG analysis. This automation represents a significant step toward advancing auditory diagnostics and improving patient outcomes.

The transfer of new insights from basic or clinical research into clinical routine is usually a lengthy and time-consuming process. Conversely, there are still many barriers to directly provide and use routine data in the context of basic and clinical research. In particular, no coherent software solution is available that allows a convenient and immediate bidirectional transfer of data between concrete treatment contexts and research settings. Here, we present a generic framework that integrates health data (e.g., clinical, molecular) and computational analytics (e.g., model predictions, statistical evaluations, visualizations) into a clinical software solution which simultaneously supports both patient-specific healthcare decisions and research efforts, while also adhering to the requirements for data protection and data quality. Specifically, our work is based on a recently established generic data management concept, for which we designed and implemented a web-based software framework that integrates data analysis, visualization as well as computer simulation and model prediction with audit trail functionality and a regulation-compliant pseudonymization service. Within the front-end application, we established two tailored views: a clinical (i.e., treatment context) perspective focusing on patient-specific data visualization, analysis and outcome prediction and a research perspective focusing on the exploration of pseudonymized data. We illustrate the application of our generic framework by two use-cases from the field of haematology/oncology. Our implementation demonstrates the feasibility of an integrated generation and backward propagation of data analysis results and model predictions at an individual patient level into clinical decision-making processes while enabling seamless integration into a clinical information system or an electronic health record.

Domänenspezifische Sprachen (DSLs) sind Software-Sprachen, die speziell für bestimmte Anwendungsdomänen entwickelt wurden. Mithilfe von DSLs können Domänenexperten ihr Domänenwissen auf einem hohen Abstraktionsniveau beschreiben. Wie andere Software-Sprachen auch, benötigen DSLs Sprachwerkzeuge, die Assistenz bei der Erstellung und Verarbeitung von domänenspezifischen Modellen bieten. Eine domänenspezifische Werkbank (DSW) ist ein Software-Werkzeug, welches mehrere solcher Sprachwerkzeuge für eine DSL miteinander integriert. Existierende Werkzeuge, die es erlauben eine DSW aufgrund der Beschreibung einer DSL automatisch generieren zu lassen, unterstützen jedoch nicht die Beschreibung und Generierung von editierbaren Sichten. Eine Sicht ist ein Teil einer DSW, der nur einen bestimmten Aspekt eines Modells darstellt. Diese Dissertation stellt spezielle Modelltransformationssprachen (MTLs) vor, mit denen die Synchronisation von Sichten in einer generierten DSW beschrieben werden kann. Dadurch können DSWs mit editierbaren Sichten mittels existierender Werkzeuge zur Generierung von Sprachwerkzeugen erstellt werden. Dafür wird eine DSW für die Nanophysik-Domäne sowie eine Taxonomie von Synchronisationstypen vorgestellt, welche es erlaubt genau zu bestimmen, welche Art von Modelltransformationen für die Synchronisation von Sichten in dieser Werkbank benötigt werden. Entsprechend dieser Anforderungen werden zwei MTLs entwickelt. Insbesondere wird eine bidirektionale MTL entwickelt. Mit solch einer Sprache kann man eine Relation, welche definiert ob zwei Modelle synchron sind, so beschreiben, dass die entsprechende Synchronisationslogik automatisch abgeleitet werden kann. Die gezeigten MTLs werden als interne DSLs - das heißt eingebettet als ausdrucksstarke Bibliotheken - in der Programmiersprache Scala implementiert. Auf diese Weise kann Scalas Typprüfung genutzt werden, um Transformationen und deren Komposition statisch zu verifizieren. / Domain-specific languages (DSLs) are software languages which are tailored to a specific application domain. DSLs enable domain experts to create domain-specific models, that is, high-level descriptions of domain knowledge. As any other software languages, DSLs rely on language tools which provide assistance for processing and managing domain-specific models. A domain-specific workbench is an integrated set of such tools for a DSL. A recently proposed approach is to automatically generate a domain-specific workbench for a DSL from a description of that DSL. However, existing tools which apply this approach do not support to describe and generate editable domain-specific views. A view is a part of domain-specific workbench that presents only one aspect of a model, for example, its hierarchical structure. This dissertation presents special model transformation languages which support the description of view synchronization in a generated domain-specific workbench. This allows a multi-view domain-specific workbench to be created with existing tools for language tool generation. We present a generated domain-specific workbench for the nanophysics domain and present a taxonomy of synchronization types. This allows us to precisely define what model transformations are required for view synchronization in that workbench. According to these requirements, we develop two transformation languages by adapting existing ones. In particular, we develop a bidirectional transformation language. With such a language one can describe a relation which defines whether two models are in sync and let the synchronization logic be inferred automatically. We implement model transformation languages as internal DSLs - that is, embedded as expressive libraries - in the Scala programming language and use Scala''s type checking for static verification of transformations and their composition.

We consider the problem of secure communications in a Gaussian two-way relay network where two nodes exchange confidential messages only via an untrusted relay. The relay is assumed to be honest but curious, i.e., an eavesdropper that conforms to the system rules and applies the intended relaying scheme. We analyze the achievable secrecy rates by applying network coding on the physical layer or the network layer and compare the results in terms of complexity, overhead, and efficiency. Further, we discuss the advantages and disadvantages of the respective approaches.

L'entraînement sans surveillance efficace et inférence dans les modèles génératifs profonds reste un problème difficile. Une approche assez simple, la machine de Helmholtz, consiste à entraîner du haut vers le bas un modèle génératif dirigé qui sera utilisé plus tard pour l'inférence approximative. Des résultats récents suggèrent que de meilleurs modèles génératifs peuvent être obtenus par de meilleures procédures d'inférence approximatives. Au lieu d'améliorer la procédure d'inférence, nous proposons ici un nouveau modèle, la machine de Helmholtz bidirectionnelle, qui garantit qu'on peut calculer efficacement les distributions de haut-vers-bas et de bas-vers-haut. Nous y parvenons en interprétant à les modèles haut-vers-bas et bas-vers-haut en tant que distributions d'inférence approximative, puis ensuite en définissant la distribution du modèle comme étant la moyenne géométrique de ces deux distributions. Nous dérivons une borne inférieure pour la vraisemblance de ce modèle, et nous démontrons que l'optimisation de cette borne se comporte en régulisateur. Ce régularisateur sera tel que la distance de Bhattacharyya sera minisée entre les distributions approximatives haut-vers-bas et bas-vers-haut. Cette approche produit des résultats de pointe en terme de modèles génératifs qui favorisent les réseaux significativement plus profonds. Elle permet aussi une inférence approximative amérliorée par plusieurs ordres de grandeur. De plus, nous introduisons un modèle génératif profond basé sur les modèles BiHM pour l'entraînement semi-supervisé. / Efficient unsupervised training and inference in deep generative models remains a challenging problem. One basic approach, called Helmholtz machine, involves training a top-down directed generative model together with a bottom-up auxiliary model used for approximate inference. Recent results indicate that better generative models can be obtained with better approximate inference procedures. Instead of improving the inference procedure, we here propose a new model, the bidirectional Helmholtz machine, which guarantees that the top-down and bottom-up distributions can efficiently invert each other. We achieve this by interpreting both the top-down and the bottom-up directed models as approximate inference distributions and by defining the model distribution to be the geometric mean of these two. We present a lower-bound for the likelihood of this model and we show that optimizing this bound regularizes the model so that the Bhattacharyya distance between the bottom-up and top-down approximate distributions is minimized. This approach results in state of the art generative models which prefer significantly deeper architectures while it allows for orders of magnitude more efficient approximate inference. Moreover, we introduce a deep generative model for semi-supervised learning problems based on BiHM models.

Conforme novas tecnologias e metodologias estão surgindo, e pesquisadores estão sedentos por ferramentas moleculares mas rápidas, mais eficientes e fáceis de usar, dominar os princípios e tecnologias do design de vetores e padronização de partes biológicas tornaram-se desafios fundamentais. Isso está abrindo espaço para o surgimento de uma disciplina inteiramente nova chamada Biologia Sintética. Esta área de estudo inovadora combina partes e módulos biológicos para criar sistemas mais confiáveis e robustos. Linhagens fúngicas são comumente alvo desses estudos, não apenas porque muitos achados fundamentais em relação à clonagem molecular surgiram das lições dadas por elas, mas também devido a um imenso e inexplorado potencial desses organismos em uma ampla gama de aplicações - desde biocombustíveis e produção de químicos finos até terapias biomédicas. Neste contexto, a presente dissertação é dividida em duas partes: a primeira diz respeito ao design e construção de uma ferramenta modular e versátil para ser aplicada em várias linhagens de fungos. Essa ferramenta é um plasmídeo binário para transformação mediada pro Agrobacterium tumefaciens, que foi construído em quatro diferentes versões contendo GFP ou mCherry como proteínas repórter e um gene sintético de resistência à higromicina como marcador de seleção. O vetor foi validado em Paracoccidioides lutzii, um patógeno oportunista humano dimórfico que é muito importante para medicina, mas ainda carecia de ferramentas genéticas eficientes. A segunda parte consiste na criação de uma biblioteca de promotores para a levedura oleaginosa Rhodosporidium toruloides, um promissor hospedeiro para a produção de bioprodutos a partir de biomassa, uma vez que pode eficientemente consumir açúcares C5 e C6 e aromáticos derivados da lignina. Vinte e nove promotores foram testados em um cassete de duplo-repórter - compreendendo ambas as proteínas fluorescentes GFP e mRuby - utilizando citometria de fluxo para análise de células únicas. A coleção de promotores apresentados neste trabalho é a maior disponível para R. toruloides até o momento e foi um avanço indispensável para superar a10 escassez de ferramentas para este organismo. Notavelmente, também apresentamos os primeiros promotores bidirecionais descritos para essa levedura e otimizamos o protocolo de transformação. Portanto, a Biologia Sintética foi eficientemente aplicada para expandir a coleção de partes biológicas padronizadas e otimizar vetores para transformação e manipulação genética de fungos. Estas ferramentas são de valor imediato e são aplicáveis a desafios muito distintos, mas igualmente importantes: a busca de novas soluções para a saúde humana e para uma economia bio-sustentável. / As new technologies and methodologies are surfacing, and researchers are now eager for fast, enhanced and easy-to-use molecular tools, mastering the principles and technologies of vector design and standardization of biological parts have become fundamental challenges. This is making room for the rise of an entirely novel discipline called Synthetic Biology. This innovative field of study combines biological parts and modules to create more reliable and robust systems. Fungal strains are commonly the target of these studies, not only because several fundamental findings regarding molecular cloning arose from lessons given by them, but also due to an immense and much unexplored potential of those organisms in a wide range of applications - ranging from biofuels and fine chemicals production to biomedical therapies. In this context, the present dissertation is divided in two parts: the first one concerns the design and construction of a modular and versatile tool to be applied in several fungal strains. This tool is a plasmid binary vector for Agrobacterium tumefaciens-mediated transformation, which was built in four different versions containing either GFP or mCherry as reporter proteins and a synthetic hygromycin resistance gene as selection marker. The vector was validated in Paracoccidioides lutzii, a dimorphic human opportunist pathogen that is very important for health care but was still lacking efficient genetic tools. The second part consists in the creation of a promoter library for the oleaginous yeast Rhodosporidium toruloides, a promising host for the production of bioproducts from biomass since it can efficiently consume C5 and C6 sugars and lignin-derived aromatics. Twenty-nine promoters were tested with a dual-reporter cassette - comprising both GFP and mRuby fluorescent proteins - using flow cytometer for single-cell analysis. The assortment of promoters presented in this work is the largest set available for R. toruloides until now and was an imperative advancement to overcome the scarcity of tools for this organism. Remarkably, we also presented the first bidirectional promoters described for this yeast and optimized the transformation protocol. Thus, we efficiently applied Synthetic Biology to expand the collection of standard biological parts and to12 optimize vectors for fungal transformation and genetic manipulation. These tools are of immediate value and are applicable for very distinct but equally important challenges: the pursuit of new solutions for human health and for a sustainable biobased economy

O presente trabalho visa o desenvolvimento de um sistema de carga e descarga de baterias de chumbo-ácido para sistemas de geração distribuída acoplada a um conjunto de painéis fotovoltaicos e conectada à rede elétrica. O conjunto de painéis opera de forma a maximizar a energia gerada através de um algoritmo MPPT (Maximum Power Point Tracking) e a rede elétrica opera como rota para o excedente produzido pelos painéis. Para que a bateria possa ser carregada e descarregada com eficiência evitando sobrecargas e descarregamentos profundos, uma metodologia de estimação do estado de carga da bateria é implementada visando um controle mais eficiente (maximizando a vida útil da bateria) e um melhor aproveitamento da energia armazenada. Para o gerenciamento da carga e descarga do banco de baterias é utilizada uma topologia bidirecional (Boost-Buck) com controlador fuzzy P+I para estabilização de corrente ou tensão. Em relação à técnica de controle para o sistema PV, um controlador PI clássico é utilizado para regular a tensão terminal e ponto de máxima potência que é definido por um algoritmo de perturbação e observação (P&O). Para validar os controladores e os modelos teóricos desenvolvidos é construído um protótipo do sistema. São analisados, também, o comportamento do sistema de carga das baterias e painel fotovoltaico sob condições extremas como queda abrupta de irradiação solar, desconexão inesperada dos painéis e mudanças no ponto ótimo de funcionamento dos mesmos. / The present work aims to develop a system of charging and discharging lead-acid batteries for distributed generation connected to a set of photovoltaic panels, and used in grid connected applications. The set of panels operates to maximize the energy produced by means of an MPPT (Maximum Power Point Tracking) algorithm and the power grid absorbs the extra power produced by the panels. For charging and discharging the battery pack and to prevent overload and deep discharge, a methodology for estimating the state of charge is implemented in order to obtain an efficient control technique (maximizing battery life), and more efficient use of the stored energy. To manage the charging and discharging of the battery bank a bidirectional topology (Boost-Buck) with fuzzy P+I controller for stabilization of current or voltage is used. To regards the PV control technique, a classical PI controller is employed to regulate the PV terminal voltage, and the maximum power point is determined by a perturb and observe (P&O) algorithm. To validate the controllers and the theoretical models developed is built a prototype system, and analyzed the behavior of the batteries charger and photovoltaic panel under extreme environmental conditions such as abrupt decrease of solar irradiation, unexpected panels disconnection and changes of the maximum power point of the panels.

O presente trabalho visa o desenvolvimento de um sistema de carga e descarga de baterias de chumbo-ácido para sistemas de geração distribuída acoplada a um conjunto de painéis fotovoltaicos e conectada à rede elétrica. O conjunto de painéis opera de forma a maximizar a energia gerada através de um algoritmo MPPT (Maximum Power Point Tracking) e a rede elétrica opera como rota para o excedente produzido pelos painéis. Para que a bateria possa ser carregada e descarregada com eficiência evitando sobrecargas e descarregamentos profundos, uma metodologia de estimação do estado de carga da bateria é implementada visando um controle mais eficiente (maximizando a vida útil da bateria) e um melhor aproveitamento da energia armazenada. Para o gerenciamento da carga e descarga do banco de baterias é utilizada uma topologia bidirecional (Boost-Buck) com controlador fuzzy P+I para estabilização de corrente ou tensão. Em relação à técnica de controle para o sistema PV, um controlador PI clássico é utilizado para regular a tensão terminal e ponto de máxima potência que é definido por um algoritmo de perturbação e observação (P&O). Para validar os controladores e os modelos teóricos desenvolvidos é construído um protótipo do sistema. São analisados, também, o comportamento do sistema de carga das baterias e painel fotovoltaico sob condições extremas como queda abrupta de irradiação solar, desconexão inesperada dos painéis e mudanças no ponto ótimo de funcionamento dos mesmos. / The present work aims to develop a system of charging and discharging lead-acid batteries for distributed generation connected to a set of photovoltaic panels, and used in grid connected applications. The set of panels operates to maximize the energy produced by means of an MPPT (Maximum Power Point Tracking) algorithm and the power grid absorbs the extra power produced by the panels. For charging and discharging the battery pack and to prevent overload and deep discharge, a methodology for estimating the state of charge is implemented in order to obtain an efficient control technique (maximizing battery life), and more efficient use of the stored energy. To manage the charging and discharging of the battery bank a bidirectional topology (Boost-Buck) with fuzzy P+I controller for stabilization of current or voltage is used. To regards the PV control technique, a classical PI controller is employed to regulate the PV terminal voltage, and the maximum power point is determined by a perturb and observe (P&O) algorithm. To validate the controllers and the theoretical models developed is built a prototype system, and analyzed the behavior of the batteries charger and photovoltaic panel under extreme environmental conditions such as abrupt decrease of solar irradiation, unexpected panels disconnection and changes of the maximum power point of the panels.

Afin de répondre aux besoins en management efficace de l’énergie électrique dans les bâtiments intelligents, le laboratoire GREMAN a proposé une nouvelle topologie d’interrupteur bidirectionnel de 600 V nommé TBBS. Les études antérieures ont validé la bidirectionnalité en courant et en tension de cette nouvelle topologie. Les travaux de recherche menées dans cette thèse avaient pour l’objectif d’approfondir, de compléter nos connaissances sur ce nouvel interrupteur bidirectionnel ainsi que sur le transistor bipolaire à fort gain. Le premier chapitre introduit le fonctionnement principal du TBBS et sa modélisation physique sous un environnement de simulation à éléments finis. Le deuxième chapitre présente le travail concernant la caractérisation expérimentale du TBBS et du transistor bipolaire à fort gain sous température contrôlée. Enfin la modélisation électrique du TBBS et du transistor bipolaire à fort gain est présentée dans le troisième et dernier chapitre. / In order to meet the requirement of more efficient electrical energy management for intelligent buildings, a new 600V bidirectional switch, named as TBBS, has been proposed by the GREMAN laboratory. Previous studies have validated the current and voltage bidirectionality of this newly proposed topology. The research work carried out in this thesis deals with a deeper and more comprehensive study of this bidirectional switch and its elementary component - the High-gain bipolar juncion transistor. The first chapter introduces the operation of the TBBS and its physical modeling in a finite element simulation environment. The second chapter presentes the research work related to the experimental caracterisation of the TBBS and the High-gain bipolar junction transistor. At last the third chapter deals with the electrical modeling of these two bipolar components.