<p dir="ltr">This research explores adversarial attacks on Graph Convolutional Transformer (GCT) models that utilize Electronic Health Record (EHR) data. As deep learning models become increasingly integral to healthcare, securing their robustness against adversarial threats is critical. This research assesses the susceptibility of GCT models to specific adversarial attacks, namely the Fast Gradient Sign Method (FGSM) and the Jacobian-based Saliency Map Attack (JSMA). It examines their effect on the model’s prediction of mortality and readmission. Through experiments conducted with the MIMIC-III and eICU datasets, the study finds that although the GCT model exhibits superior performance in processing EHR data under normal conditions, its accuracy drops when subjected to adversarial conditions—from an accuracy of 86% with test data to about 57% and an area under the curve (AUC) from 0.86 to 0.51. These findings averaged across both datasets and attack methods, underscore the urgent need for effective adversarial defense mechanisms in AI systems used in healthcare. This thesis contributes to the field by identifying vulnerabilities and suggesting various strategies to enhance the resilience of GCT models against adversarial manipulations.</p>

Urgent scientifically-informed action is needed to stabilise the Earth System amidst anthropogenic climate change. Particularly, the notable transgression of the ‘biosphere integrity’ Planetary Boundary needs to be addressed. Modern Earth System Models struggle to accurately represent regional to local-scale climate features and biodiversity aspects. Recent developments allow to tackle these issues using Artificial Intelligence. This dissertation focuses on two main aspects: (i) deriving high spatio-temporal resolution climate data from coarser models; and (ii) integrating high-temporal-resolution climate data into Species Distribution Models. Three specific objectives were defined: Obj1 Improving Perfect Prognosis – Statistical Downscaling methods through modern Deep Learning algorithms. Obj2 Downscaling a high-resolution multivariate climate ensemble. Obj3 Employ the resulting dataset to improve Species Distribution Models’ projections. The objectives are connected to the three articles that support this cumulative dissertation. Its scope is limited to the Free State of Saxony, Germany, where local high-resolution climate data and high-quality observations of endangered vascular plant species were employed. From a broader perspective, these efforts should contribute to the overarching goal of bridging the gap between the scales of species distribution and climate models while establishing open-source, reproducible, and scalable containerised frameworks. Recent Deep Learning algorithms were leveraged to accomplish (i). The proposed frameworks enhance previous performance of Perfect Prognosis – Statistical Downscaling approaches, while ensuring repeatability. The key near-surface variables considered are precipitation, water vapour pressure, radiation, wind speed, and, maximum, mean and minimum temperature. The assumptions that support the Perfect Prognosis approach were thoroughly examined, confirming the robustness of the methods. The downscaled ensemble exhibits a novel output resolution of daily 1 km, which can serve as input for multiple climate impact studies, especially for local-scale decision-making and in topographically complex regions. Considerable methodological implementations were proposed and thoroughly analysed to achieve (ii). Despite notable limitations, Species Distribution Models are frequently used in climate change conservation planning. Thus, recent developments in climate data resolution could improve their usefulness and reliability, which have been previously constraint to coarse temporal aggregates in the projection domain. The presented framework provides fine-grained species suitability projections and satisfactory spatio-temporal transferability, albeit worrying trends. These improved projections are a step forward towards tailored conservation efforts. Limitations of Machine Learning methods and Species Distribution Models are addressed. Substantial avenues for future improvements are thoroughly discussed. As results suggest further reduction of suitable habitats, yet another call for swift action towards low-carbon societies is made. This requires maximising climate change mitigation and adaptation measures, along with a swift transition from short-term profit-driven policies to long-term sustainable development, but primarily, a collective shift in consciousness from anthropocentric positions to ecocentric policies and societies.:Contents Declaration of conformity........................................................ I Abstract....................................................................... III Zusammenfassung.................................................................. V Resumen........................................................................ VII Acknowledgments................................................................. IX List of Figures................................................................. XV List of Tables................................................................. XIX Symbols and Acronyms........................................................... XXI I Prelude & Foundations 1 1 Introduction................................................................... 3 1.1 Motivation – Human Impact on Earth....................................... 3 1.2 Earth System Modelling and Downscaling................................... 5 1.3 Biosphere’s Response to Recent Changes................................... 8 1.4 Species Distribution Models.............................................. 9 1.5 Objectives.............................................................. 10 1.6 Scope................................................................... 10 1.7 Outline................................................................. 10 2 Methodological Basis.......................................................... 13 2.1 Introduction to Artificial Intelligence Methods......................... 13 2.1.1 Artificial Intelligence........................................... 13 2.1.2 Machine Learning.................................................. 14 2.1.3 Deep Learning..................................................... 14 2.2 Downscaling Techniques.................................................. 15 2.2.1 Dynamical Downscaling............................................. 15 2.2.2 Statistical Downscaling........................................... 15 2.2.2.1 Model Output Statistics................................... 16 2.2.2.2 Perfect Prognosis......................................... 16 2.3 Species Distribution Models: Temporal Aspects........................... 17 2.4 Computational Framework................................................. 18 2.4.1 High-Performance Computing........................................ 18 2.4.2 Containers........................................................ 18 2.5 Remarks on Reproducibility.............................................. 19 II Articles’ Synthesis 21 3 Data.......................................................................... 23 3.1 Study Area.............................................................. 23 3.2 ReKIS................................................................... 24 3.3 ERA5.................................................................... 24 3.4 CORDEX.................................................................. 24 3.5 Species Occurrences..................................................... 25 3.6 WorldClim............................................................... 26 4 Methodological Implementations................................................ 27 4.1 Advancing Statistical Downscaling....................................... 27 4.1.1 Transfer Function Calibration.................................... 27 4.1.2 Evaluation....................................................... 29 4.1.3 Repeatability.................................................... 29 4.2 Downscaling a Multivariate Ensemble..................................... 30 4.2.1 Transfer Function Adaptations.................................... 30 4.2.2 Validation....................................................... 30 4.2.3 Perfect Prognosis Assumptions Evaluation......................... 31 4.3 Integrating High-Temporal-Resolution into SDMs.......................... 32 4.3.1 Climate Data..................................................... 32 4.3.1.1 Predictor Sets.......................................... 32 4.3.1.2 Temporal Approaches..................................... 33 4.3.2 SDM Implementation............................................... 33 4.3.3 Spatio-Temporal Thinning & Trimming.............................. 33 4.3.4 Meta-analysis.................................................... 34 4.3.5 Pseudo-Reality Assessment........................................ 34 4.3.6 Spatio-Temporal Transferability.................................. 34 5 Results & Discussions......................................................... 35 5.1 Advancing Statistical Downscaling....................................... 35 5.1.1 Performance Improvement.......................................... 35 5.1.2 Repeatability.................................................... 36 5.1.3 Transfer Function Suitability.................................... 38 5.2 Downscaling a Multivariate Ensemble..................................... 39 5.2.1 Transfer Function performance.................................... 39 5.2.2 Bias-Correction.................................................. 40 5.2.3 Pseudo-Reality................................................... 42 5.2.4 Projections...................................................... 43 5.3 Integrating High-Temporal-Resolution into SDMs.......................... 45 5.3.1 Predictor Set Evaluation for H2k................................. 45 5.3.2 Temporal Approach Comparison..................................... 46 5.3.3 Spatio-Temporal Transferability.................................. 47 5.3.4 Suitability Projections.......................................... 47 III Insights 51 6 Summary....................................................................... 53 6.1 Article A1.............................................................. 53 6.2 Article A2.............................................................. 54 6.3 Article A3.............................................................. 56 7 Conclusions and Outlook....................................................... 59 References 65 Articles 81 A1 Repeatable high-resolution statistical downscaling through deep learning..... 83 A2 Downscaling CORDEX Through Deep Learning to Daily 1 km Multivariate Ensemble in Complex Terrain............................................................. 103 A3 Integrating High-Temporal-Resolution Climate Projections into Species Distribu- tion Model..................................................................... 127 / Um das Erdsystem angesichts des anthropogenen Klimawandels zu stabilisieren, sind Maßnahmen auf Basis wissenschaftlicher Erkenntnisse dringend erforderlich. Insbesondere muss die drastisch Überschreitung der planetaren Grenze ‘Integrität der Biosphäre’ angegangen werden. Bisher haben aber Modelle des Erdsystems Schwierigkeiten, regionale bis lokale Klimamerkmale und Aspekte der Biodiversität genau abzubilden. Aktuelle Entwicklungen ermöglichen es, diese Herausforderungen mithilfe von Künstlicher Intelligenz anzugehen. Diese Dissertation konzentriert sich auf zwei Hauptaspekte: (i) die Ableitung von Klimadaten mit hoher räumlicher und zeitlicher Auflösung aus groberen Modellen und (ii) die Integration von Klimadaten mit hoher zeitlicher Auflösung in Modelle zur Artverbreitung. Es wurden drei konkrete Ziele definiert: Ziel1 Verbesserung von Perfect Prognosis – Statistische Downscaling-Methoden durch moderne Deep Learning-Algorithmen Ziel2 Downscaling eines hochauflösenden multivariaten Klimaensembles Ziel3 Verwendung des resultierenden Datensatzes zur Verbesserung von Prognosen in Modellen zur Artverbreitung Diese Ziele werden in drei wissenschaftlichen Artikeln beantwortet, auf die diese kumulative Dissertation sich stützt. Der Anwendungsbereich erstreckt sich auf den Freistaat Sachsen, Deutschland, wo lokale hochauflösende Klimadaten und hochwertige Beobachtungen gefährdeter Gefäßpflanzenarten verwendet wurden. In einer breiteren Perspektive tragen diese Bemühungen dazu bei, die Kluft zwischen regionalen sowie zeitlichen Skalen der Artverbreitung und Klimamodellen zu überbrücken und gleichzeitig Open-Source-, reproduzierbare und skalierbare containerisierte Frameworks zu etablieren. Aktuelle Deep Learning-Algorithmen wurden eingesetzt, um Hauptaspekt (i) zu erreichen. Die vorgeschlagenen Frameworks verbessern die bisherige Leistung von Perfect Prognosis – Statistische Downscaling-Ansätzen und gewährleisten gleichzeitig die Wiederholbarkeit. Die wichtigsten bodennahen Variablen, die berücksichtigt werden, sind Niederschlag, Wasserdampfdruck, Strahlung, Windgeschwindigkeit sowie Maximal-, Durchschnitts- und Minimaltemperatur. Die Annahmen, die den Perfect Prognosis-Ansatz unterstützen, wurden analysiert und bestätigen die Robustheit der Methoden. Das downscaled Ensemble weist eine neuartige Auflösung von 1 km auf Tagesbasis auf, welches als Grundlage für mehrere Studien zu den Auswirkungen des Klimawandels dienen kann, insbesondere für Entscheidungsfindung auf lokaler Ebene und in topografisch komplexen Regionen. Es wurden umfassende methodische Implementierungen vorgeschlagen und analysiert, um Hauptaspekt (ii) zu erreichen. Trotz großer Einschränkungen werden Modelle zur Artverbreitung häufig in der Klimaschutzplanung eingesetzt. Daher könnten aktuelle Entwicklungen in der Klimadatenauflösung deren Nützlichkeit und Zuverlässigkeit verbessern, die bisher auf grobe zeitliche Aggregatformen im Projektionsbereich beschränkt waren. Das vorgestellte Framework bietet feingliedrige Prognosen zur Eignung von Arten und zufriedenstellende räumlich-zeitliche Übertragbarkeit, trotz besorgniserregender Trends. Diese verbesserten Prognosen sind ein Schritt in Richtung maßgeschneiderter Naturschutzmaßnahmen. Einschränkungen von Machine Learning-Methoden und Modellen zur Artverbreitung werden untersucht. Substanzielle Möglichkeiten zur zukünftigen Verbesserung werden ausführlich erörtert. Da die Ergebnisse darauf hinweisen, dass geeignete Lebensräume weiter abnehmen, wird erneut zum schnellen Handeln in Richtung kohlenstoffarmer Gesellschaften aufgerufen. Dies erfordert die Maximierung von Maßnahmen zur Bekämpfung des Klimawandels und zur Anpassung, zusammen mit einem raschen Übergang von kurzfristig Profitorientierten Politiken zu langfristiger nachhaltiger Entwicklung, aber vor allem zu einem kollektiven Bewusstseinswandel von anthropozentrischen Positionen zu ökozentrischen Politiken und Gesellschaften.:Contents Declaration of conformity........................................................ I Abstract....................................................................... III Zusammenfassung.................................................................. V Resumen........................................................................ VII Acknowledgments................................................................. IX List of Figures................................................................. XV List of Tables................................................................. XIX Symbols and Acronyms........................................................... XXI I Prelude & Foundations 1 1 Introduction................................................................... 3 1.1 Motivation – Human Impact on Earth....................................... 3 1.2 Earth System Modelling and Downscaling................................... 5 1.3 Biosphere’s Response to Recent Changes................................... 8 1.4 Species Distribution Models.............................................. 9 1.5 Objectives.............................................................. 10 1.6 Scope................................................................... 10 1.7 Outline................................................................. 10 2 Methodological Basis.......................................................... 13 2.1 Introduction to Artificial Intelligence Methods......................... 13 2.1.1 Artificial Intelligence........................................... 13 2.1.2 Machine Learning.................................................. 14 2.1.3 Deep Learning..................................................... 14 2.2 Downscaling Techniques.................................................. 15 2.2.1 Dynamical Downscaling............................................. 15 2.2.2 Statistical Downscaling........................................... 15 2.2.2.1 Model Output Statistics................................... 16 2.2.2.2 Perfect Prognosis......................................... 16 2.3 Species Distribution Models: Temporal Aspects........................... 17 2.4 Computational Framework................................................. 18 2.4.1 High-Performance Computing........................................ 18 2.4.2 Containers........................................................ 18 2.5 Remarks on Reproducibility.............................................. 19 II Articles’ Synthesis 21 3 Data.......................................................................... 23 3.1 Study Area.............................................................. 23 3.2 ReKIS................................................................... 24 3.3 ERA5.................................................................... 24 3.4 CORDEX.................................................................. 24 3.5 Species Occurrences..................................................... 25 3.6 WorldClim............................................................... 26 4 Methodological Implementations................................................ 27 4.1 Advancing Statistical Downscaling....................................... 27 4.1.1 Transfer Function Calibration.................................... 27 4.1.2 Evaluation....................................................... 29 4.1.3 Repeatability.................................................... 29 4.2 Downscaling a Multivariate Ensemble..................................... 30 4.2.1 Transfer Function Adaptations.................................... 30 4.2.2 Validation....................................................... 30 4.2.3 Perfect Prognosis Assumptions Evaluation......................... 31 4.3 Integrating High-Temporal-Resolution into SDMs.......................... 32 4.3.1 Climate Data..................................................... 32 4.3.1.1 Predictor Sets.......................................... 32 4.3.1.2 Temporal Approaches..................................... 33 4.3.2 SDM Implementation............................................... 33 4.3.3 Spatio-Temporal Thinning & Trimming.............................. 33 4.3.4 Meta-analysis.................................................... 34 4.3.5 Pseudo-Reality Assessment........................................ 34 4.3.6 Spatio-Temporal Transferability.................................. 34 5 Results & Discussions......................................................... 35 5.1 Advancing Statistical Downscaling....................................... 35 5.1.1 Performance Improvement.......................................... 35 5.1.2 Repeatability.................................................... 36 5.1.3 Transfer Function Suitability.................................... 38 5.2 Downscaling a Multivariate Ensemble..................................... 39 5.2.1 Transfer Function performance.................................... 39 5.2.2 Bias-Correction.................................................. 40 5.2.3 Pseudo-Reality................................................... 42 5.2.4 Projections...................................................... 43 5.3 Integrating High-Temporal-Resolution into SDMs.......................... 45 5.3.1 Predictor Set Evaluation for H2k................................. 45 5.3.2 Temporal Approach Comparison..................................... 46 5.3.3 Spatio-Temporal Transferability.................................. 47 5.3.4 Suitability Projections.......................................... 47 III Insights 51 6 Summary....................................................................... 53 6.1 Article A1.............................................................. 53 6.2 Article A2.............................................................. 54 6.3 Article A3.............................................................. 56 7 Conclusions and Outlook....................................................... 59 References 65 Articles 81 A1 Repeatable high-resolution statistical downscaling through deep learning..... 83 A2 Downscaling CORDEX Through Deep Learning to Daily 1 km Multivariate Ensemble in Complex Terrain............................................................. 103 A3 Integrating High-Temporal-Resolution Climate Projections into Species Distribu- tion Model..................................................................... 127 / Acción urgente científicamente informada es necesaria para estabilizar el sistema terrestre en medio del cambio climático antropogénico. En particular, la notable transgresión del límite planetario de ’integridad de la biosfera’ debe abordarse. Los modernos modelos del sistema terrestre tienen dificultades para representar con precisión las características climáticas a escala regional y local, así como los aspectos de la biodiversidad. Desarrollos recientes permiten abordar estos problemas mediante la inteligencia artificial. Esta disertación se enfoca en dos aspectos principales: (i) derivar datos climáticos de alta resolución espacio-temporal a partir de modelos más gruesos; y (ii) integrar datos climáticos de alta resolución temporal en modelos de distribución de especies. Se definieron tres objetivos específicos: Obj1 Mejorar los métodos de pronóstico perfecto – reducción de escala estadística mediante algoritmos modernos de aprendizaje profundo. Obj2 Generar un conjunto climático multivariado de alta resolución. Obj3 Emplear el conjunto de datos resultante para mejorar las proyecciones de los modelos de distribución de especies. Los objetivos están vinculados a los tres artículos que respaldan esta disertación acumulativa. Su alcance se limita al Estado Libre de Sajonia, Alemania, donde se emplearon datos climáticos locales de alta resolución y observaciones de alta calidad de especies de plantas vasculares en peligro de extinción. Desde una perspectiva más amplia, estos esfuerzos deberían contribuir a la meta general de cerrar la brecha entre las escalas de la distribución de especies y los modelos climáticos, mientras que se establecen marcos de trabajo contenedorizados de código abierto, reproducibles y escalables. Algoritmos recientes de aprendizaje profundo fueron aprovechados para lograr (i). Los marcos de trabajo propuestos mejoran el rendimiento previo de los métodos de pronóstico perfecto – reducción de escala estadística, al tiempo que garantizan la repetibilidad. Las variables clave de la superficie cercana consideradas son precipitación, presión de vapor de agua, radiación, velocidad del viento, así como la temperatura máxima, media y mínima. Se examinaron meticulosamente las suposiciones que respaldan el método de pronóstico perfecto, confirmando la robustez de las propuestas. El conjunto reducido de escala exhibe una novedosa resolución diaria de 1 km, el cual puede servir como insumo para múltiples estudios de impacto climático, especialmente para la toma de decisiones a nivel local y en regiones topográficamente complejas. Se propusieron y analizaron minuciosamente considerables implementaciones metodológicas para lograr (ii). A pesar de sus notables limitaciones, los modelos de distribución de especies son utilizados con frecuencia en la planificación de la conservación debido al cambio climático. Por lo tanto, los desarrollos recientes en la resolución de datos climáticos podrían mejorar su utilidad y confiabilidad, ya que antes se limitaban a agregados temporales gruesos en el caso de las proyecciones. El marco de trabajo presentado proporciona proyecciones de idoneidad de especies detalladas y una transferibilidad espacio-temporal satisfactoria, aunque con tendencias preocupantes. Estas proyecciones mejoradas son un paso adelante en los esfuerzos de conservación a la medida. Se abordan las limitaciones de los métodos de aprendizaje automático y de los modelos de distribución de especies. Se discuten a fondo posibilidades sustanciales para futuras mejoras. Dado que los resultados sugieren una mayor reducción de hábitats adecuados, se hace otro llamado a la acción rápida hacia sociedades bajas en carbono. Esto requiere maximizar las medidas de mitigación y adaptación al cambio climático, junto con una transición rápida de políticas orientadas a beneficios a corto plazo hacia un desarrollo sostenible a largo plazo, pero principalmente, un cambio colectivo de conciencia, desde posiciones antropocéntricas hacia políticas y sociedades ecocéntricas.:Contents Declaration of conformity........................................................ I Abstract....................................................................... III Zusammenfassung.................................................................. V Resumen........................................................................ VII Acknowledgments................................................................. IX List of Figures................................................................. XV List of Tables................................................................. XIX Symbols and Acronyms........................................................... XXI I Prelude & Foundations 1 1 Introduction................................................................... 3 1.1 Motivation – Human Impact on Earth....................................... 3 1.2 Earth System Modelling and Downscaling................................... 5 1.3 Biosphere’s Response to Recent Changes................................... 8 1.4 Species Distribution Models.............................................. 9 1.5 Objectives.............................................................. 10 1.6 Scope................................................................... 10 1.7 Outline................................................................. 10 2 Methodological Basis.......................................................... 13 2.1 Introduction to Artificial Intelligence Methods......................... 13 2.1.1 Artificial Intelligence........................................... 13 2.1.2 Machine Learning.................................................. 14 2.1.3 Deep Learning..................................................... 14 2.2 Downscaling Techniques.................................................. 15 2.2.1 Dynamical Downscaling............................................. 15 2.2.2 Statistical Downscaling........................................... 15 2.2.2.1 Model Output Statistics................................... 16 2.2.2.2 Perfect Prognosis......................................... 16 2.3 Species Distribution Models: Temporal Aspects........................... 17 2.4 Computational Framework................................................. 18 2.4.1 High-Performance Computing........................................ 18 2.4.2 Containers........................................................ 18 2.5 Remarks on Reproducibility.............................................. 19 II Articles’ Synthesis 21 3 Data.......................................................................... 23 3.1 Study Area.............................................................. 23 3.2 ReKIS................................................................... 24 3.3 ERA5.................................................................... 24 3.4 CORDEX.................................................................. 24 3.5 Species Occurrences..................................................... 25 3.6 WorldClim............................................................... 26 4 Methodological Implementations................................................ 27 4.1 Advancing Statistical Downscaling....................................... 27 4.1.1 Transfer Function Calibration.................................... 27 4.1.2 Evaluation....................................................... 29 4.1.3 Repeatability.................................................... 29 4.2 Downscaling a Multivariate Ensemble..................................... 30 4.2.1 Transfer Function Adaptations.................................... 30 4.2.2 Validation....................................................... 30 4.2.3 Perfect Prognosis Assumptions Evaluation......................... 31 4.3 Integrating High-Temporal-Resolution into SDMs.......................... 32 4.3.1 Climate Data..................................................... 32 4.3.1.1 Predictor Sets.......................................... 32 4.3.1.2 Temporal Approaches..................................... 33 4.3.2 SDM Implementation............................................... 33 4.3.3 Spatio-Temporal Thinning & Trimming.............................. 33 4.3.4 Meta-analysis.................................................... 34 4.3.5 Pseudo-Reality Assessment........................................ 34 4.3.6 Spatio-Temporal Transferability.................................. 34 5 Results & Discussions......................................................... 35 5.1 Advancing Statistical Downscaling....................................... 35 5.1.1 Performance Improvement.......................................... 35 5.1.2 Repeatability.................................................... 36 5.1.3 Transfer Function Suitability.................................... 38 5.2 Downscaling a Multivariate Ensemble..................................... 39 5.2.1 Transfer Function performance.................................... 39 5.2.2 Bias-Correction.................................................. 40 5.2.3 Pseudo-Reality................................................... 42 5.2.4 Projections...................................................... 43 5.3 Integrating High-Temporal-Resolution into SDMs.......................... 45 5.3.1 Predictor Set Evaluation for H2k................................. 45 5.3.2 Temporal Approach Comparison..................................... 46 5.3.3 Spatio-Temporal Transferability.................................. 47 5.3.4 Suitability Projections.......................................... 47 III Insights 51 6 Summary....................................................................... 53 6.1 Article A1.............................................................. 53 6.2 Article A2.............................................................. 54 6.3 Article A3.............................................................. 56 7 Conclusions and Outlook....................................................... 59 References 65 Articles 81 A1 Repeatable high-resolution statistical downscaling through deep learning..... 83 A2 Downscaling CORDEX Through Deep Learning to Daily 1 km Multivariate Ensemble in Complex Terrain............................................................. 103 A3 Integrating High-Temporal-Resolution Climate Projections into Species Distribu- tion Model..................................................................... 127

Ett vanligt problem inom supervised learning är brist på taggad träningsdata. Naive semi-supervised deep learning är en träningsteknik som ämnar att mildra detta problem genom att generera pseudo-taggad data och därefter låta ett neuralt nätverk träna på denna samt en mindre mängd taggad data. Detta arbete undersöker om denna teknik kan förbättras genom användandet av röstning. Flera neurala nätverk tränas genom den framtagna tekniken, naive semi-supervised deep learning eller supervised learning och deras träffsäkerhet utvärderas därefter. Resultaten visade nästan enbart försämringar då röstning användes. Dock verkar inte förutsättningarna för röstning ha varit särskilt goda, vilket gör det svårt att dra en säker slutsats kring effekterna av röstning. Även om röstning inte gav förbättringar har NSSDL visat sig vara mycket effektiv. Det finns flera applikationsområden där tekniken i framtiden skulle kunna användas med goda resultat.

<div>Fluorescence microscopy has emerged as a powerful tool for studying cell biology because it enables the acquisition of 3D image volumes deeper into tissue and the imaging of complex subcellular structures. Fluorescence microscopy images are frequently distorted by motion resulting from animal respiration and heartbeat which complicates the quantitative analysis of biological structures needed to characterize the structure and constituency of tissue volumes. This thesis describes a two pronged approach to quantitative analysis consisting of non-rigid registration and deep convolutional neural network segmentation. The proposed image registration method is capable of correcting motion artifacts in three dimensional fluorescence microscopy images collected over time. In particular, our method uses 3D B-Spline based nonrigid registration using a coarse-to-fine strategy to register stacks of images collected at different time intervals and 4D rigid registration to register 3D volumes over time. The results show that the proposed method has the ability of correcting global motion artifacts of sample tissues in four dimensional space, thereby revealing the motility of individual cells in the tissue.</div><div><br></div><div>We describe in thesis nuclei segmentation methods using deep convolutional neural networks, data augmentation to generate training images of different shapes and contrasts, a refinement process combining segmentation results of horizontal, frontal, and sagittal planes in a volume, and a watershed technique to enumerate the nuclei. Our results indicate that compared to 3D ground truth data, our method can successfully segment and count 3D nuclei. Furthermore, a microscopy image synthesis method based on spatially constrained cycle-consistent adversarial networks is used to efficiently generate training data. A 3D modified U-Net network is trained with a combination of Dice loss and binary cross entropy metrics to achieve accurate nuclei segmentation. A multi-task U-Net is utilized to resolve overlapping nuclei. This method was found to achieve high accuracy object-based and voxel-based evaluations.</div>

Submitted by Angela Maria de Oliveira (amolivei@uepg.br) on 2019-02-28T17:58:29Z No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) Renan Rosa.pdf: 4067327 bytes, checksum: eb0bd9e84fbd89a24b4a397c9655fa62 (MD5) / Made available in DSpace on 2019-02-28T17:58:29Z (GMT). No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) Renan Rosa.pdf: 4067327 bytes, checksum: eb0bd9e84fbd89a24b4a397c9655fa62 (MD5) Previous issue date: 2018-11-19 / As pragas em lavouras causam prejuízos econômicos na agricultura, reduzindo a produção e consequentemente os lucros. O manejo de pragas é essencial, para reduzir estes prejuízos, e consiste na identificação e posterior controle desse tipo de ameaça. O controle é fundamentalmente dependente da identificação, pois é a partir dela que o manejo é feito. A identificação é feita visualmente, baseando-se nas características da praga. Essas características são inerentes e diferem de espécie para espécie. Devido à dificuldade da identificação, esse processo é realizado principalmente por profissionais especializados na área, o que acarreta na concentração do conhecimento. Esta dissertação apresenta uma metodologia para classificação de pragas por meio de técnicas de computação, onde um sistema computacional do tipo clienteservidor foi criado a fim de prover a classificação de pragas por meio de serviço, que é realizado pelo uso de rede neural convolucional baseada na arquitetura Inception V3. As pragas Anticarsia Gemmatalis, Helicoverpa armigera e Spodoptera Cosmioides, foram escolhidas para classificação por serem bastante comuns no estado do Paraná. A rede neural convolucional obteve índice de acerto de 92,5%. / Pests on crops cause economic damage to agriculture, reducing production and consequently profits. Pest management is essential to reduce these losses, and consists in the identification and subsequent control of this type of threat. Control is fundamentally dependent on identification, because management is done from it. The identification is made visually, based on the characteristics of the pest. These characteristics are inherent and differ from species to species. Due to the difficulty of identification, this process is carried out mainly by professionals specialized in the area, which entails the concentration of knowledge. This dissertation presents a methodology for pest classification by means of computational techniques, in which a client-server computational system was created in order to provide pest classification by means of a service, which is performed by the use of convolutional neural network based in the Inception V3 architecture. The pests Anticarsia Gemmatalis, Helicoverpa armigera and Spodoptera Cosmioides, were chosen for classification because they are quite common in the state of Paraná. The convolutional neural network obtained a success rate of 92.5%.

Les travaux de recherche de cette thèse consistent à mettre en place des solutions performantes et légères permettant de répondre aux problèmes de sécurisation de produits sensibles. Motivé par une collaboration avec différents acteurs au sein du projet Nuc-Track,le développement d'un système de sécurité biométrique, possiblement multimodal, mènera à une étude sur différentes caractéristiques biométriques telles que le visage, les empreintes digitales et le réseau vasculaire. Cette thèse sera axée sur une adéquation algorithme et architecture, dans le but de minimiser la taille de stockage des modèles d'apprentissages tout en garantissant des performances optimales. Cela permettra leur stockage sur un support personnel, respectant ainsi les normes de vie privée. / Research of this thesis consists in setting up efficient and light solutions to answer the problems of securing sensitive products. Motivated by a collaboration with various stakeholders within the Nuc-Track project, the development of a biometric security system, possibly multimodal, will lead to a study on various biometric features such as the face, fingerprints and the vascular network. This thesis will focus on an algorithm and architecture matching, with the aim of minimizing the storage size of the learning models while guaranteeing optimal performances. This will allow it to be stored on a personal support, thus respecting privacy standards.

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2018-11-21T11:05:44Z No. of bitstreams: 2 Dissertação - Vinícius Araújo Santos - 2018.pdf: 14601054 bytes, checksum: e02a8bcd3cdc93bf2bf202c3933b3f27 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2018-11-21T11:06:26Z (GMT) No. of bitstreams: 2 Dissertação - Vinícius Araújo Santos - 2018.pdf: 14601054 bytes, checksum: e02a8bcd3cdc93bf2bf202c3933b3f27 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2018-11-21T11:06:26Z (GMT). No. of bitstreams: 2 Dissertação - Vinícius Araújo Santos - 2018.pdf: 14601054 bytes, checksum: e02a8bcd3cdc93bf2bf202c3933b3f27 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2018-10-11 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Visual Odometry is an important process in image based navigation of robots. The standard methods of this field rely on the good feature matching between frames where feature detection on images stands as a well adressed problem within Computer Vision. Such techniques are subject to illumination problems, noise and poor feature localization accuracy. Thus, 3D information on a scene may mitigate the uncertainty of the features on images. Deep Learning techniques show great results when dealing with common difficulties of VO such as low illumination conditions and bad feature selection. While Visual Odometry and Deep Learning have been connected previously, no techniques applying Siamese Convolutional Networks on depth infomation given by disparity maps have been acknowledged as far as this work’s researches went. This work aims to fill this gap by applying Deep Learning to estimate egomotion through disparity maps on an Siamese architeture. The SiameseVO-Depth architeture is compared to state of the art techniques on OV by using the KITTI Vision Benchmark Suite. The results reveal that the chosen methodology succeeded on the estimation of Visual Odometry although it doesn’t outperform the state-of-the-art techniques. This work presents fewer steps in relation to standard VO techniques for it consists of an end-to-end solution and demonstrates a new approach of Deep Learning applied to Visual Odometry. / Odometria Visual é um importante processo na navegação de robôs baseada em imagens. Os métodos clássicos deste tema dependem de boas correspondências de características feitas entre imagens sendo que a detecção de características em imagens é um tema amplamente discutido no campo de Visão Computacional. Estas técnicas estão sujeitas a problemas de iluminação, presença de ruído e baixa de acurácia de localização. Nesse contexto, a informação tridimensional de uma cena pode ser uma forma de mitigar as incertezas sobre as características em imagens. Técnicas de Deep Learning têm demonstrado bons resultados lidando com problemas comuns em técnicas de OV como insuficiente iluminação e erros na seleção de características. Ainda que já existam trabalhos que relacionam Odometria Visual e Deep Learning, não foram encontradas técnicas que utilizem Redes Convolucionais Siamesas com sucesso utilizando informações de profundidade de mapas de disparidade durante esta pesquisa. Este trabalho visa preencher esta lacuna aplicando Deep Learning na estimativa do movimento por de mapas de disparidade em uma arquitetura Siamesa. A arquitetura SiameseVO-Depth proposta neste trabalho é comparada à técnicas do estado da arte em OV utilizando a base de dados KITTI Vision Benchmark Suite. Os resultados demonstram que através da metodologia proposta é possível a estimativa dos valores de uma Odometria Visual ainda que o desempenho não supere técnicas consideradas estado da arte. O trabalho proposto possui menos etapas em comparação com técnicas clássicas de OV por apresentar-se como uma solução fim-a-fim e apresenta nova abordagem no campo de Deep Learning aplicado à Odometria Visual.

[pt] A Extração e Atribuição de Citações é a tarefa de identificar citações de um texto e associá-las a seus autores. Neste trabalho, apresentamos um sistema de Extração e Atribuição de Citações para a língua portuguesa. A tarefa de Extração e Atribuição de Citações foi abordada anteriormente utilizando diversas técnicas e para uma variedade de linguagens e datasets. Os modelos tradicionais para a tarefa consistem em extrair manualmente um rico conjunto de atributos e usá-los para alimentar um classificador raso. Neste trabalho, ao contrário da abordagem tradicional, evitamos usar atributos projetados à mão, usando técnicas de aprendizagem não supervisionadas e redes neurais profundas para automaticamente aprender atributos relevantes para resolver a tarefa. Ao evitar a criação manual de atributos, nosso modelo de aprendizagem de máquina tornou-se facilmente adaptável a outros domínios e linguagens. Nosso modelo foi treinado e avaliado no corpus GloboQuotes e sua métrica de desempenho F1 é igual a 89.43 por cento. / [en] Quotation Extraction and Attribution is the task of identifying quotations from a given text and associating them to their authors. In this work, we present a Quotation Extraction and Attribution system for the Portuguese language. The Quotation Extraction and Attribution task has been previously approached using various techniques and for a variety of languages and datasets. Traditional models to this task consist of extracting a rich set of hand-designed features and using them to feed a shallow classifier. In this work, unlike the traditional approach, we avoid using hand-designed features using unsupervised learning techniques and deep neural networks to automatically learn relevant features to solve the task. By avoiding design features by hand, our machine learning model became easily adaptable to other languages and domains. Our model is trained and evaluated at the GloboQuotes corpus, and its F1 performance metric is equal to 89.43 percent.

La reconnaissance des opinions d'un locuteur dans une interaction orale est une étape cruciale pour améliorer la communication entre un humain et un agent virtuel. Dans cette thèse, nous nous situons dans une problématique de traitement automatique de la parole (TAP) sur les phénomènes d'opinions dans des interactions orales spontanées naturelles. L'analyse d'opinion est une tâche peu souvent abordée en TAP qui se concentrait jusqu'à peu sur les émotions à l'aide du contenu vocal et non verbal. De plus, la plupart des systèmes récents existants n'utilisent pas le contexte interactionnel afin d'analyser les opinions du locuteur. Dans cette thèse, nous nous penchons sur ces sujet. Nous nous situons dans le cadre de la détection automatique en utilisant des modèles d’apprentissage statistiques. Après une étude sur la modélisation de la dynamique de l'opinion par un modèle à états latents à l’intérieur d'un monologue, nous étudions la manière d’intégrer le contexte interactionnel dialogique, et enfin d'intégrer l'audio au texte avec différents types de fusion. Nous avons travaillé sur une base de données de Vlogs au niveau d'un sentiment global, puis sur une base de données d'interactions dyadiques multimodales composée de conversations ouvertes, au niveau du tour de parole et de la paire de tours de parole. Pour finir, nous avons fait annoté une base de données en opinion car les base de données existantes n'étaient pas satisfaisantes vis-à-vis de la tâche abordée, et ne permettaient pas une comparaison claire avec d'autres systèmes à l'état de l'art.A l'aube du changement important porté par l’avènement des méthodes neuronales, nous étudions différents types de représentations: les anciennes représentations construites à la main, rigides mais précises, et les nouvelles représentations apprises de manière statistique, générales et sémantiques. Nous étudions différentes segmentations permettant de prendre en compte le caractère asynchrone de la multi-modalité. Dernièrement, nous utilisons un modèle d'apprentissage à états latents qui peut s'adapter à une base de données de taille restreinte, pour la tâche atypique qu'est l'analyse d'opinion, et nous montrons qu'il permet à la fois une adaptation des descripteurs du domaine écrit au domaine oral, et servir de couche d'attention via son pouvoir de clusterisation. La fusion multimodale complexe n'étant pas bien gérée par le classifieur utilisé, et l'audio étant moins impactant sur l'opinion que le texte, nous étudions différentes méthodes de sélection de paramètres pour résoudre ces problèmes. / 2588/5000Recognizing a speaker's opinions in an oral interaction is a crucial step in improving communication between a human and a virtual agent. In this thesis, we find ourselves in a problematic of automatic speech processing (APT) on opinion phenomena in natural spontaneous oral interactions. Opinion analysis is a task that is not often addressed in TAP that focused until recently on emotions using voice and non-verbal content. In addition, most existing legacy systems do not use the interactional context to analyze the speaker's opinions. In this thesis, we focus on these topics.We are in the context of automatic detection using statistical learning models. A study on modeling the dynamics of opinion by a model with latent states within a monologue, we study how to integrate the context interactional dialogical, and finally to integrate audio to text with different types of fusion. We worked on a basic Vlogs data at a global sense, and on the basis of multimodal data dyadic interactions composed of open conversations, at the turn of speech and word pair of towers. Finally, we annotated database in opinion because existing database were not satisfactory vis-à-vis the task addressed, and did not allow a clear comparison with other systems in the state art.At the dawn of significant change brought by the advent of neural methods, we study different types of representations: the ancient representations built by hand, rigid, but precise, and new representations learned statistically, and general semantics. We study different segmentations to take into account the asynchronous nature of multi-modality. Recently, we are using a latent state learning model that can adapt to a small database, for the atypical task of opinion analysis, and we show that it allows both an adaptation of the descriptors of the written domain to the oral domain, and serve as an attention layer via its clustering power. Complex multimodal fusion is not well managed by the classifier used, and audio being less impacting on opinion than text, we study different methods of parameter selection to solve these problems.

Die Anwendung von Deep Learning in der manuellen Montage birgt großes Potenzial, Montagezeiten zu reduzieren und Montagefehler zu vermeiden. Indem der Montageablauf mithilfe einer Kamera erfasst und die aufgezeichneten Bilder durch einen Objekterkennungsalgorithmus analysiert werden, lassen sich Position, Lage und Art der montierten Bauteile bestimmen. Daraus lassen sich wiederum Informationen über Arbeitsschritte, Montagefehler oder den aktuellen Zustand des Produkts ableiten, sodass die Mitarbeiter bei der Montage durch entsprechende Anweisungen unterstützt werden können. Es stellt sich jedoch die Frage, inwieweit gegenwärtige Produkte für den Einsatz von Deep Learning geeignet sind. Nur wenn die zu montierenden Bauteile sicher erkannt werden, ist der Einsatz in der manuellen Montage sinnvoll. Bestehende Gestaltungsrichtlinien adressieren diesen Aspekt bislang nicht. Im Forschungsprojekt wurde daher untersucht, welche Eigenschaften Produkte aufweisen sollten, um eine optimale Objekterkennung zu ermöglichen. Dazu wurden Hypothesen zu positiven und negativen Bauteileigenschaften hinsichtlich der Erkennungsgenauigkeit formuliert und in praktischen Versuchen überprüft. Dabei konnte gezeigt werden, dass alle untersuchten Objekte durch den eingesetzten Objekterkennungsalgorithmus sehr gut detektiert werden. Aus den vorliegenden Forschungsergebnissen lassen sich daher keine Einschränkungen in der Produktgestaltung ableiten.