Tesis por compendio / This dissertation focused in the comprehensive analysis of the capabilities of some non-tested types of Artificial Neural Networks, specifically: the Probabilistic Neural Networks (PNN) and the Multi-Layer Perceptron (MLP) Ensembles. The analysis of the capabilities of these techniques was performed using the native brown trout (Salmo trutta; Linnaeus, 1758), the bermejuela (Achondrostoma arcasii; Robalo, Almada, Levy & Doadrio, 2006) and the redfin barbel (Barbus haasi; Mertens, 1925) as target species. The analyses focused in the predictive capabilities, the interpretability of the models and the effect of the excess of zeros in the training datasets, which for presence-absence models is directly related to the concept of data prevalence (i.e. proportion of presence instances in the training dataset). Finally, the effect of the spatial scale (i.e. micro-scale or microhabitat scale and meso-scale) in the habitat suitability models and consequently in the e-flow assessment was studied in the last chapter. / Esta tesis se centra en el análisis comprensivo de las capacidades de algunos tipos de Red Neuronal Artificial aún no testados: las Redes Neuronales Probabilísticas (PNN) y los Conjuntos de Perceptrones Multicapa (MLP Ensembles). Los análisis sobre las capacidades de estas técnicas se desarrollaron utilizando la trucha común (Salmo trutta; Linnaeus, 1758), la bermejuela (Achondrostoma arcasii; Robalo, Almada, Levy & Doadrio, 2006) y el barbo colirrojo (Barbus haasi; Mertens, 1925) como especies nativas objetivo. Los análisis se centraron en la capacidad de predicción, la interpretabilidad de los modelos y el efecto del exceso de ceros en las bases de datos de entrenamiento, la así llamada prevalencia de los datos (i.e. la proporción de casos de presencia sobre el conjunto total). Finalmente, el efecto de la escala (micro-escala o escala de microhábitat y meso-escala) en los modelos de idoneidad del hábitat y consecuentemente en la evaluación de caudales ambientales se estudió en el último capítulo. / Aquesta tesis se centra en l'anàlisi comprensiu de les capacitats d'alguns tipus de Xarxa Neuronal Artificial que encara no han estat testats: les Xarxes Neuronal Probabilístiques (PNN) i els Conjunts de Perceptrons Multicapa (MLP Ensembles). Les anàlisis sobre les capacitats d'aquestes tècniques es varen desenvolupar emprant la truita comuna (Salmo trutta; Linnaeus, 1758), la madrilla roja (Achondrostoma arcasii; Robalo, Almada, Levy & Doadrio, 2006) i el barb cua-roig (Barbus haasi; Mertens, 1925) com a especies objecte d'estudi. Les anàlisi se centraren en la capacitat predictiva, interpretabilitat dels models i en l'efecte de l'excés de zeros a la base de dades d'entrenament, l'anomenada prevalença de les dades (i.e. la proporció de casos de presència sobre el conjunt total). Finalment, l'efecte de la escala (micro-escala o microhàbitat i meso-escala) en els models d'idoneïtat de l'hàbitat i conseqüentment en l'avaluació de cabals ambientals es va estudiar a l'últim capítol. / Muñoz Mas, R. (2016). Multivariate approaches in species distribution modelling: Application to native fish species in Mediterranean Rivers [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/76168 / Compendio

<p dir="ltr">Autonomous vehicles have gained substantial traction as the future of transportation, necessitating continuous research and innovation. While 2D object detection and instance segmentation methods have made significant strides, 3D object detection offers unparalleled precision. Deep neural network-based 3D object detection, coupled with sensor fusion, has become indispensable for self-driving vehicles, enabling a comprehensive grasp of the spatial geometry of physical objects. In our study of a Lidar-based 3D object detection network using point clouds, we propose a novel architectural model based on You Only Look Once (YOLO) framework. This innovative model combines the efficiency and accuracy of the YOLOv8 network, a swift 2D standard object detector, and a state-of-the-art model, with the real-time 3D object detection capability of the Complex YOLO model. By integrating the YOLOv8 model as the backbone network and employing the Euler Region Proposal (ERP) method, our approach achieves rapid inference speeds, surpassing other object detection models while upholding high accuracy standards. Our experiments, conducted on the KITTI dataset, demonstrate the superior efficiency of our new architectural model. It outperforms its predecessors, showcasing its prowess in advancing the field of 3D object detection in autonomous vehicles.</p>

Severe flooding and excess nutrient pollution, exacerbated by heightened anthropogenic pressures (e.g., climate change, urbanization, land use change, unsustainable agricultural practices), have been detrimental to riverine systems and their estuaries. The degradation of riverine systems can negatively impact human and environmental health, as well as local, regional, and even global economies. Floods provide beneficial ecosystem services (e.g., processing pollutants, transferring nutrients and sediment, supporting biodiversity), but they can also damage infrastructure and result in the loss of human life. Meanwhile, eutrophication can cause anoxic dead zones, harming aquatic ecosystems and public health. To address the issues facing riverine systems, focus has shifted to watershed-scale management plans. However, it can prove challenging to quantify the cumulative impacts of multiple stream restoration projects within a single watershed on flooding and nutrient removal. Previous studies have quantified the effects of stream restoration on flood attenuation. However, our first study fills a substantial knowledge gap by evaluating the impacts of different floodplain restoration practices, varied by location and length, on flood attenuation and floodplain inundation dynamics at the watershed scale during more frequent storm recurrence intervals (i.e., 2-year, 1-year, 0.5-year, and monthly). We created a 1D HEC-RAS model to simulate the effects of Stage 0 restoration within a 4th-order generic watershed based on the Chesapeake Bay watershed. By varying the percent river length restored and location, we found that Stage 0 restoration, especially in 2nd-order rivers, can be particularly effective at enhancing flood attenuation and floodplain inundation locally and farther downstream. We addressed the water quality component by using a random forest machine learning approach coupled with artificial neural networks to find trends and predict nitrate removal rates associated with spatial, temporal, hydrologic, and restoration features. Our results showed that hydrologic conditions were the most important variable for predicting actual nitrate removal rates. Overall, both studies demonstrate the importance of hydrologic connectivity for flood attenuation, channel-floodplain exchange, and nutrient processing. / Maryland Department of Natural Resources; National Fish and Wildlife Foundation through the U.S. Environmental Protection Agency’s Chesapeake Bay Program Office; Chesapeake Bay Trust / Master of Science / Severe flooding and nutrient pollution from sources such as urban and agricultural runoff have been detrimental to the health of rivers. The degradation of rivers can negatively impact human and environmental health, as well as local, regional, and even global economies. Floods can be both helpful, by providing water quality benefits and supporting wildlife, and harmful, causing damage and even loss of life. Excess nutrients, such as nitrogen, can create underwater zones void of life, with serious consequences for aquatic life and public health. To address the flooding and water quality issues facing rivers, focus has shifted to landscapelevel river network management plans. However, it can prove challenging to understand the impacts of multiple stream restoration projects within a larger river network on flooding and nutrient removal. We address the flooding component by using a model to simulate the effects of different floodplain restoration techniques on a medium-sized watershed that is generally based on streams that flow into the Chesapeake Bay. Our model simulated small, relatively frequent storm events that, on average, occur every two years to once a month. By varying restoration length and location, we found that restoration practices with lower streambanks can be particularly effective at slowing down floods, reducing their overall severity by allowing more water to access the floodplains. This was especially true when restoration occurred in smaller streams, and the effects were seen both locally and farther downstream. We address the water quality component by using a different model to find patterns and predict nutrient removal rates associated with different landscape, seasonal, storm event, and restoration features. Our results showed that the most important variable for predicting nutrient removal rates was whether a stream was experiencing normal flow or stormflow conditions. Overall, both studies demonstrate the importance of restoring rivers in a manner that encourages water to flow from the channel into the floodplains during smaller storm events, because this will reduce the severity of downstream flooding while simultaneously improving water quality.

The development of new solutions incorporating artificial intelligence (AI) within the medical field is an area of great interest. However, access to comprehensive and diverse datasets is restricted due to the sensitive nature of the data. A potential solution to this is to generatesynthetic datasets based on real medical data. Synthetic data could protect the integrity of the subjects while preserving the inherent information necessary for training AI models and be generated in greater quantity than otherwise available. This thesis project aims to generate reliable clinical trial subject data using a generative adversarial network (GAN). The main data set used is a mock clinical trial dataset consisting of multiple subject visits, however an additional data set containing authentic medical data is also used for better insights into the model’s ability to learn underlying relationships. The thesis also investigates training strategies for simulating the temporal dimension and the missing values in the data. The GAN model used is an altered version of the Conditional Tabular GAN (CTGAN)made to be compatible with the preprocessed clinical trial mock data, and multiple model architectures and number of training epochs are examined. The results show great potential for GAN models on clinical trial datasets, especially for real-life data. One model, trained on the authentic dataset, generates near-perfect synthetic data with respect to column distributions and correlation between columns. The results also show that classification models trained on synthetic data and tested on real data have the potential to match the performance of classification models trained on real data. While the synthetic data replicates the missing values, no definitive conclusion can be drawn regarding the temporal characteristics due to the sparsity of the mock dataset and lack of real correlations in it. Although the results are promising, further experiments on authentic datasets with less sparsity are required.

The use of audio and video communication has increased exponentially over the last decade and has gone from speech over GSM to HD resolution video conference between continents on mobile devices. As the use becomes more widespread the interest in delivering high quality media increases even on devices with limited resources. This includes both development and enhancement of the communication chain but also the topic of objective measurements of the perceived quality. The focus of this thesis work has been to perform enhancement within speech encoding and video decoding, to measure influence factors of audio and video performance, and to build methods to predict the perceived video quality. The audio enhancement part of this thesis addresses the well known problem in the GSM system with an interfering signal generated by the switching nature of TDMA cellular telephony. Two different solutions are given to suppress such interference internally in the mobile handset. The first method involves the use of subtractive noise cancellation employing correlators, the second uses a structure of IIR notch filters. Both solutions use control algorithms based on the state of the communication between the mobile handset and the base station. The video enhancement part presents two post-filters. These two filters are designed to improve visual quality of highly compressed video streams from standard, block-based video codecs by combating both blocking and ringing artifacts. The second post-filter also performs sharpening. The third part addresses the problem of measuring audio and video delay as well as skewness between these, also known as synchronization. This method is a black box technique which enables it to be applied on any audiovisual application, proprietary as well as open standards, and can be run on any platform and over any network connectivity. The last part addresses no-reference (NR) bitstream video quality prediction using features extracted from the coded video stream. Several methods have been used and evaluated: Multiple Linear Regression (MLR), Artificial Neural Network (ANN), and Least Square Support Vector Machines (LS-SVM), showing high correlation with both MOS and objective video assessment methods as PSNR and PEVQ. The impact from temporal, spatial and quantization variations on perceptual video quality has also been addressed, together with the trade off between these, and for this purpose a set of locally conducted subjective experiments were performed.

Cette thèse explore la reconnaissance de gestes à partir de capteurs inertiels pour Smartphone. Ces gestes consistent en la réalisation d'un tracé dans l'espace présentant une valeur sémantique, avec l'appareil en main. Notre étude porte en particulier sur l'apprentissage de métrique entre signatures gestuelles grâce à l'architecture "Siamoise" (réseau de neurones siamois, SNN), qui a pour but de modéliser les relations sémantiques entre classes afin d'extraire des caractéristiques discriminantes. Cette architecture est appliquée au perceptron multicouche (MultiLayer Perceptron). Les stratégies classiques de formation d'ensembles d'apprentissage sont essentiellement basées sur des paires similaires et dissimilaires, ou des triplets formés d'une référence et de deux échantillons respectivement similaires et dissimilaires à cette référence. Ainsi, nous proposons une généralisation de ces approches dans un cadre de classification, où chaque ensemble d'apprentissage est composé d’une référence, un exemple positif, et un exemple négatif pour chaque classe dissimilaire. Par ailleurs, nous appliquons une régularisation sur les sorties du réseau au cours de l'apprentissage afin de limiter les variations de la norme moyenne des vecteurs caractéristiques obtenus. Enfin, nous proposons une redéfinition du problème angulaire par une adaptation de la notion de « sinus polaire », aboutissant à une analyse en composantes indépendantes non-linéaire supervisée. A l'aide de deux bases de données inertielles, la base MHAD (Multimodal Human Activity Dataset) ainsi que la base Orange, composée de gestes symboliques inertiels réalisés avec un Smartphone, les performances de chaque contribution sont caractérisées. Ainsi, des protocoles modélisant un monde ouvert, qui comprend des gestes inconnus par le système, mettent en évidence les meilleures capacités de détection et rejet de nouveauté du SNN. En résumé, le SNN proposé permet de réaliser un apprentissage supervisé de métrique de similarité non-linéaire, qui extrait des vecteurs caractéristiques discriminants, améliorant conjointement la classification et le rejet de gestes inertiels. / As consumer devices become more and more ubiquitous, new interaction solutions are required. In this thesis, we explore inertial-based gesture recognition on Smartphones, where gestures holding a semantic value are drawn in the air with the device in hand. In our research, speed and delay constraints required by an application are critical, leading us to the choice of neural-based models. Thus, our work focuses on metric learning between gesture sample signatures using the "Siamese" architecture (Siamese Neural Network, SNN), which aims at modelling semantic relations between classes to extract discriminative features, applied to the MultiLayer Perceptron. Contrary to some popular versions of this algorithm, we opt for a strategy that does not require additional parameter fine tuning, namely a set threshold on dissimilar outputs, during training. Indeed, after a preprocessing step where the data is filtered and normalised spatially and temporally, the SNN is trained from sets of samples, composed of similar and dissimilar examples, to compute a higher-level representation of the gesture, where features are collinear for similar gestures, and orthogonal for dissimilar ones. While the original model already works for classification, multiple mathematical problems which can impair its learning capabilities are identified. Consequently, as opposed to the classical similar or dissimilar pair; or reference, similar and dissimilar sample triplet input set selection strategies, we propose to include samples from every available dissimilar classes, resulting in a better structuring of the output space. Moreover, we apply a regularisation on the outputs to better determine the objective function. Furthermore, the notion of polar sine enables a redefinition of the angular problem by maximising a normalised volume induced by the outputs of the reference and dissimilar samples, which effectively results in a Supervised Non-Linear Independent Component Analysis. Finally, we assess the unexplored potential of the Siamese network and its higher-level representation for novelty and error detection and rejection. With the help of two real-world inertial datasets, the Multimodal Human Activity Dataset as well as the Orange Dataset, specifically gathered for the Smartphone inertial symbolic gesture interaction paradigm, we characterise the performance of each contribution, and prove the higher novelty detection and rejection rate of our model, with protocols aiming at modelling unknown gestures and open world configurations. To summarise, the proposed SNN allows for supervised non-linear similarity metric learning, which extracts discriminative features, improving both inertial gesture classification and rejection.

Depuis son introduction en 1982, la carte auto-organisatrice de Kohonen (Self-Organizing Map : SOM) a prouvé ses capacités de classification et visualisation des données multidimensionnelles dans différents domaines d’application. Les implémentations matérielles de la carte SOM, en exploitant le taux de parallélisme élevé de l’algorithme de Kohonen, permettent d’augmenter les performances de ce modèle neuronal souvent au détriment de la flexibilité. D’autre part, la flexibilité est offerte par les implémentations logicielles qui quant à elles ne sont pas adaptées pour les applications temps réel à cause de leurs performances temporelles limitées. Dans cette thèse nous avons proposé une architecture matérielle distribuée, adaptable, flexible et extensible de la carte SOM à base de NoC dédiée pour une implantation matérielle sur FPGA. A base de cette approche, nous avons également proposé une architecture matérielle innovante d’une carte SOM à structure croissante au cours de la phase d’apprentissage / Since its introduction in 1982, Kohonen’s Self-Organizing Map (SOM) showed its ability to classify and visualize multidimensional data in various application fields. Hardware implementations of SOM, by exploiting the inherent parallelism of the Kohonen algorithm, allow to increase the overall performances of this neuronal network, often at the expense of the flexibility. On the other hand, the flexibility is offered by software implementations which on their side are not suited for real-time applications due to the limited time performances. In this thesis we proposed a distributed, adaptable, flexible and scalable hardware architecture of SOM based on Network-on-Chip (NoC) designed for FPGA implementation. Moreover, based on this approach we also proposed a novel hardware architecture of a growing SOM able to evolve its own structure during the learning phase

Orientador: Alexandre César Rodrigues da Silva / Resumo: O reconhecimento de atividades humanas abrange diversas técnicas de classificação que permitem identificar padrões específicos do comportamento humano no momento da ocorrência. A identificação é realizada analisando dados gerados por diversos sensores corporais, entre os quais destaca-se o acelerômetro, pois responde tanto à frequência como à intensidade dos movimentos. A identificação de atividades é uma área bastante explorada. Porém, existem desafios que necessitam ser superados, podendo-se mencionar a necessidade de sistemas leves, de fácil uso e aceitação por parte dos usuários e que cumpram com requerimentos de consumo de energia e de processamento de grandes quantidades de dados. Neste trabalho apresenta-se o desenvolvimento do Sistema de Reconhecimento de atividades Humanas e Avaliação do Equilíbrio Postural, denominado SIRAH. O sistema está baseado no uso de um acelerômetro localizado na cintura do usuário. As duas fases do reconhecimento de atividades são apresentadas, fase Offline e fase Online. A fase Offline trata do treinamento de uma rede neural artificial do tipo perceptron de três camadas. No treinamento foram avaliados três estudos de caso com conjuntos de atributos diferentes, visando medir o desempenho do classificador na diferenciação de 3 posturas e 4 atividades. No primeiro caso o treinamento foi realizado com 15 atributos, gerados no domínio do tempo, com os que a rede neural artificial alcançou uma precisão de 94,40%. No segundo caso foram gerados 34 ... (Resumo completo, clicar acesso eletrônico abaixo) / Doutor

As redes ópticas passivas (PON), em virtude da oferta de maior largura de banda a custos relativamente baixos, vêm se destacando como possível candidata para suprir a demanda dos novos serviços como, tráfego de voz, vídeo, dados e de serviços móveis, exigidos pelos usuários finais. Uma importante candidata, para realizar o controle de acesso nas PONs, é a técnica de acesso múltiplo por divisão de código óptico (OCDMA), por apresentar características relevantes, como maior segurança e capacidade flexível sob demanda. No entanto, agentes físicos externos, como as variações de temperatura ambiental no enlace, exercem uma influência considerável sobre as condições de operação das redes ópticas. Especificamente, nas OCDMA-PONs, os efeitos da variação de temperatura ambiental no enlace de transmissão, afetam o valor do pico do autocorrelação do código OCDMA a ser detectado, degradando a qualidade de serviço (QoS), além do aumento da taxa de erro de bit (BER) do sistema. O presente trabalho apresenta duas novas propostas de técnicas, utilizando sistemas inteligentes, mais precisamente, controladores lógicos fuzzy (FLC) aplicados nos transmissores e nos receptores das OCDMA-PONs, com o objetivo de mitigar os efeitos de variação de temperatura. Os resultados das simulações mostram que o desempenho da rede é melhorado quando as abordagens propostas são empregadas. Por exemplo, para a distância de propagação de 10 km e variações de temperatura de 20°C, o sistema com FLC, suporta 40 usuários simultâneos com a BER = 10-9, enquanto que, sem FLC, acomoda apenas 10. Ainda neste trabalho, é proposta uma nova técnica de classificação de códigos OCDMA, com o uso de redes neurais artificiais, mais precisamente, mapas auto-organizáveis de Kohonen (SOM), importante para que o sistema de gerenciamento da rede possa oferecer uma maior segurança para os usuários. Por fim, sem o uso de técnica inteligente, é apresentada, uma nova proposta de código OCDMA, cujo formalismo desenvolvido, permite generalizar a obtenção de código com propriedades distintas, como diversas ponderações e comprimentos de códigos. / Passive optical networks (PON), due to the provision of higher bandwidth at relatively low cost, have been excelling as a possible candidate to meet the demand of new services, such as voice traffic, video, data and mobile services, as required by end users. An important candidate to perform access control in PONs, is the Optical Code-Division Multiple-Access (OCDMA) technique, due to relevant characteristics, such as improved security and flexible capacity on demand. However, external physical agents, such as variations in environmental temperature on the Fiber Optic Link, have considerable influence on the operating conditions of optical networks. Specifically, in OCDMA-PONs, the effects of environmental temperature variation in the transmission link affect the peak value on the autocorrelation of the OCDMA code to be detected, degrading the quality of service (QoS), in addition to increasing the Bit Error Rate (BER) of the system. This thesis presents two new proposals of techniques using intelligent systems, more precisely, Fuzzy Logic Controllers (FLC) applied on the transmitters and receivers of OCDMA-PONs, in order to mitigate the effects of temperature variation. The simulation results show that the network performance is improved when the proposed approaches are employed. For example, for the propagation distance of 10 kilometers and temperature variations of 20°C, the FLC system supports 40 simultaneous users at BER = 10-9, whereas without the FLC, the system can accommodate only 10. Furthermore, in this work is proposed a new technique of OCDMA codes classification, using Artificial Neural Networks (ANN), more precisely, the Self-Organizing Maps (SOM) of Kohonen, important for the network management system to provide increased security for users. Finally, without the use of intelligent technique, it is presented a new proposal of OCDMA code, whose formalism developed, allows to generalize the code acquisition with distinct properties, such as different weights and length codes.

No description available.