Implementação de uma matriz de neurônios dinamicamente reconfigurável para descrição de topologias de redes neurais artificiais multilayer perceptrons

The Artificial Neural Networks (ANN), which is one of the branches of Artificial Intelligence (AI), are being employed as a solution to many complex problems existing in several areas. To solve these problems, it is essential that its implementation is done in hardware. Among the strategies to be adopted and met during the design phase and implementation of RNAs in hardware, connections between neurons are the ones that need more attention. Recently, are RNAs implemented both in application specific integrated circuits's (Application Specific Integrated Circuits - ASIC) and in integrated circuits configured by the user, like the Field Programmable Gate Array (FPGA), which have the ability to be partially rewritten, at runtime, forming thus a system Partially Reconfigurable (SPR), the use of which provides several advantages, such as flexibility in implementation and cost reduction. It has been noted a considerable increase in the use of FPGAs for implementing ANNs. Given the above, it is proposed to implement an array of reconfigurable neurons for topologies Description of artificial neural network multilayer perceptrons (MLPs) in FPGA, in order to encourage feedback and reuse of neural processors (perceptrons) used in the same area of the circuit. It is further proposed, a communication network capable of performing the reuse of artificial neurons. The architecture of the proposed system will configure various topologies MLPs networks through partial reconfiguration of the FPGA. To allow this flexibility RNAs settings, a set of digital components (datapath), and a controller were developed to execute instructions that define each topology for MLP neural network.

Estratégias distribuídas para a coexistência de redes de banda larga sem fio em bandas não-licensiadas

The increasing demand for Internet data traffic in wireless broadband access networks requires both the development of efficient, novel wireless broadband access technologies and the allocation of new spectrum bands for that purpose. The introduction of a great number of small cells in cellular networks allied to the complimentary adoption of Wireless Local Area Network (WLAN) technologies in unlicensed spectrum is one of the most promising concepts to attend this demand. One alternative is the aggregation of Industrial, Science and Medical (ISM) unlicensed spectrum to licensed bands, using wireless networks defined by Institute of Electrical and Electronics Engineers (IEEE) and Third Generation Partnership Project (3GPP). While IEEE 802.11 (Wi-Fi) networks are aggregated to Long Term Evolution (LTE) small cells via LTE / WLAN Aggregation (LWA), in proposals like Unlicensed LTE (LTE-U) and LWA the LTE air interface itself is used for transmission on the unlicensed band. Wi-Fi technology is widespread and operates in the same 5 GHz ISM spectrum bands as the LTE proposals, which may bring performance decrease due to the coexistence of both technologies in the same spectrum bands. Besides, there is the need to improve Wi-Fi operation to support scenarios with a large number of neighbor Overlapping Basic Subscriber Set (OBSS) networks, with a large number of Wi-Fi nodes (i.e. dense deployments). It is long known that the overall Wi-Fi performance falls sharply with the increase of Wi-Fi nodes sharing the channel, therefore there is the need for introducing mechanisms to increase its spectral efficiency. This work is dedicated to the study of coexistence between different wireless broadband access systems operating in the same unlicensed spectrum bands, and how to solve the coexistence problems via distributed coordination mechanisms. The problem of coexistence between different networks (i.e. LTE and Wi-Fi) and the problem of coexistence between different networks of the same technology (i.e. multiple Wi-Fi OBSSs) is analyzed both qualitatively and quantitatively via system-level simulations, and the main issues to be faced are identified from these results. From that, distributed coordination mechanisms are proposed and evaluated via system-level simulations, both for the inter-technology coexistence problem and intra-technology coexistence problem. Results indicate that the proposed solutions provide significant gains when compare to the situation without distributed coordination.

Utilização da transformada de wavelet para detecção de variações anormais de tensão em sistemas elétricos com geração distribuída

The increasing demand in electricity and decrease forecast, increasingly, of fossil fuel reserves, as well as increasing environmental concern in the use of these have generated a concern about the quality of electricity generation, making it well welcome new investments in generation through alternative, clean and renewable sources. Distributed generation is one of the main solutions for the independent and selfsufficient generating systems, such as the sugarcane industry. This sector has grown considerably, contributing expressively in the production of electricity to the distribution networks. Faced with this situation, one of the main objectives of this study is to propose the implementation of an algorithm to detect islanding disturbances in the electrical system, characterized by situations of under- or overvoltage. The algorithm should also commonly quantize the time that the system was operating in these conditions, to check the possible consequences that will be caused in the electric power system. In order to achieve this it used the technique of wavelet multiresolution analysis (AMR) for detecting the generated disorders. The data obtained can be processed so as to be used for a possible predictive maintenance in the protection equipment of electrical network, since they are prone to damage on prolonged operation under abnormal conditions of frequency and voltage.

Classificação de nódulos pulmonares em imagens tomográficas utilizando redes neurais artificiais em cascata

Lung cancer is the most common of malignant tumors, with 1.59 million new cases worldwide in 2012. Early detection is the main factor to determine the survival of patients affected by this disease. Furthermore, the correct classification is important to define the most appropriate therapeutic approach as well as suggest the prognosis and the clinical disease evolution. Among the exams used to detect lung cancer, computed tomography have been the most indicated. However, CT images are naturally complex and even experts medical are subject to fault detection or classification. In order to assist the detection of malignant tumors, computer-aided diagnosis systems have been developed to aid reduce the amount of false positives biopsies. In this work it was developed an automatic classification system of pulmonary nodules on CT images by using Artificial Neural Networks. Morphological, texture and intensity attributes were extracted from lung nodules cut tomographic images using elliptical regions of interest that they were subsequently segmented by Otsu method. These features were selected through statistical tests that compare populations (T test of Student and U test of Mann-Whitney); from which it originated a ranking. The features after selected, were inserted in Artificial Neural Networks (backpropagation) to compose two types of classification; one to classify nodules in malignant and benign (network 1); and another to classify two types of malignancies (network 2); featuring a cascade classifier. The best networks were associated and its performance was measured by the area under the ROC curve, where the network 1 and network 2 achieved performance equal to 0.901 and 0.892 respectively.

Diseño e implementación de una herramienta de visualización para análisis en tiempo real de redes SDN/OpenFlow

Las Redes Definidas por Software (Software Defined Networking) permiten la monitorización y el control centralizado de la red, de forma que los administradores pueden tener una visión real y completa de la misma. El análisis y visualización de los diferentes parámetros obtenidos representan la forma más viable y práctica de programar la red en función de las necesidades del usuario. Por este motivo, en este proyecto se desarrolla una arquitectura modular cuyo objetivo es presentar en tiempo real la información que se monitoriza en una red SDN. En primera instancia, las diferentes métricas monitorizadas (error, retardo y tasa de datos) son almacenadas en una base de datos, para que en una etapa posterior se realice el análisis de dichas métricas. Finalmente, los resultados obtenidos, tanto de métricas en tiempo real como de los datos estadísticos, son presentados en una aplicación web. La información es obtenida a través de la interfaz REST que expone el controlador Floodlight y para el análisis de la información se plantea una comparación entre los valores medios y máximos del conjunto de datos. Los resultados obtenidos muestran gráficamente de forma clara y precisa las diferentes métricas de monitorización. Además, debido al carácter modular de la arquitectura, se ofrece un valor añadido a los sistemas actuales de monitorización SDN.

Relatório de Estágio na Inficon AG e na PT Inovação & Sistemas

This document focuses the projects developed during two independent internships, which were carried out at Inficon AG and PT Inovação & Sistemas. Since the research areas of both internships are unrelated, individual abstracts are presented.

Utilização de dispositivos de armazenamento de energia em apoio ao funcionamento de redes isoladas

Temos assistido nos últimos anos à proliferação da produção distribuída de electricidade, sobretudo, com recurso a fontes de energia renováveis, implicando a natural reestruturação das redes eléctricas existentes, desde a produção até ao consumidor final. As constantes preocupações na garantia da qualidade de serviço e até mesmo em termos ambientais, levam a que a operacionalidade das redes seja cada vez mais eficiente, visando a integração de tecnologias emergentes como é o caso dos sistemas de armazenamento de energia. A aposta nas energias de origem renovável nomeadamente a solar e a eólica, representa uma forma cada vez mais presente de geração de electricidade, tendo como grande inconveniente o regime de intermitência a que estão sujeitas, não se conseguindo tirar proveitos absolutos de todas as potencialidades que estas fontes proporcionam. Existem actualmente sistemas de armazenamento de energia que permitem optimizar o comportamento das redes. Nesta dissertação é feita uma abordagem a alguns desses sistemas, tendo como objectivo principal a demonstração das potencialidades de optimização dos sistemas de produção e distribuição de energia eléctrica com recurso a sistemas de armazenamento de energia, em redes isoladas e interligadas. É também feito um estudo do comportamento dinâmico de uma rede com vários cenários de ocorrência de defeitos, com e sem armazenamento de energia. Para isso a base deste trabalho consistiu na familiarização com uma ferramenta de grande potencial na simulação dinâmica de redes eléctricas, utilizado por prestigiados grupos de energia a nível mundial, na qual foi implementada a rede de teste e efectuadas as simulações do estudo.

Sistemas curriculares para la gestión de información y conocimiento institucional. Estudio de caso

La organización del conocimiento en el contexto de las Ciencias de la Información tiene como esencia la información y el conocimiento debidamente documentado o registrado. La organización del conocimiento como proceso, envuelve tanto la descripción física como de los contenidos de los objetos informacionales. Y el producto de ese proceso descriptivo es la representación de los atributos de un objeto o conjunto de objetos. Las representaciones son construidas por lenguajes elaborados específicamente para los objetivos de la organización en los sistemas de información. Lenguajes que se subdividen en lenguajes que describen el documento (el soporte físico del objeto) y lenguajes que describen la información (los contenidos).A partir de esta premisa la siguiente investigación tiene como objetivo general analizarlos sistemas de Gestión de Información y Conocimiento Institucional principalmente los que proponen utilizar el Currículum Vitae del profesor como única fuente de información, medición y representación de la información y el conocimiento de una organización. Dentro delos principales resultados se muestra la importancia de usar el currículo personal como fuente de información confiable y normalizada; una síntesis de los principales sistemas curriculares que existen a nivel internacional y regional; así como el gráfico del modelo de datos del caso de estudio; y por último, la propuesta del uso de las ontologías como principal herramienta para la organización semántica de la información en un sistema de gestión de información y conocimiento.

Análisis de los efectos del problema del nodo oculto en las redes IEEE 802.15.7. Propuesta de solución

Programa de doctorado: Tecnologías de la Información y sus Aplicaciones. La fecha de publicación es la fecha de lectura

Estudios de predicción en series temporales de datos meteorológicos utilizando redes neuronales recurrentes

[ES] En los últimos años, en el campo de las energías renovables, la energía eólica ha sido una de las que mas se ha desarrollado e invertido. La importancia de las predicciones de viento radica en la ayuda que aportan para planificar y anticiparse a los valores futuros que afectarán al sistema, ayudando a gestionar la adquisición de los recursos necesarios con antelación suficiente. Recientemente se han desarrollado nuevas arquitecturas de redes recurrentes que resultan muy prometedoras para realizar predicción. En este trabajo se probará y experimentará con dichas arquitecturas para realizar distintas predicciones de la velocidad del viento en un horizonte de corto y muy corto plazo a partir de datos de series temporales de viento.

Control de interrupciones de vídeo streaming móvil en arquitecturas Android usando técnicas de realidad aumentada y WebRTC

Programa de doctorado: Tecnología de la Información y sus aplicaciones. La fecha de publicación es la fecha de lectura

Aplicación de redes neuronales para la resolución de problemas complejos de confiabilidad y riesgo

Programa de doctorado: Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería. La fecha de publicación es la fecha de lectura

Enquadramento cartográfico do Complexo de Granja (NW Ceará, Brasil) e avaliação de sistemas de referenciação geográfica

Este estudo destaca os benefícios da análise metódica da cartografia de base que suporta a produção de Cartografia Geológica. Em certas regiões, as cartas base publicadas estão ainda associadas a redes geodésicas clássicas e, com frequência, são introduzidos erros quando se desconsideram parâmetros essenciais como a Projecção Cartográfica e o Datum Geodésico. Com o uso sistemático dos dispositivos de GPS e dos Sistemas de Informação Geográfica para a elaboração das cartas geológicas, é imprescindível o conhecimento prévio do Sistema de Coordenadas ao qual devem estar ajustados os dados geo-espaciais. Neste estudo de caso, as diferenças e os erros associados à aquisição de coordenadas entre os Data geocêntricos WGS84 e SIRGAS2000 são residuais, atendendo aos parâmetros das cartas base, à região do globo, o campo de acção e a escala, minimizando assim a propagação de erros de posicionamento e georreferenciação subsequentes.

Previsão do índice bovespa por meio de redes neurais artificiais: uma análise comparada aos métodos tradicionais de séries de tempo

Forecast is the basis for making strategic, tactical and operational business decisions. In financial economics, several techniques have been used to predict the behavior of assets over the past decades.Thus, there are several methods to assist in the task of time series forecasting, however, conventional modeling techniques such as statistical models and those based on theoretical mathematical models have produced unsatisfactory predictions, increasing the number of studies in more advanced methods of prediction. Among these, the Artificial Neural Networks (ANN) are a relatively new and promising method for predicting business that shows a technique that has caused much interest in the financial environment and has been used successfully in a wide variety of financial modeling systems applications, in many cases proving its superiority over the statistical models ARIMA-GARCH. In this context, this study aimed to examine whether the ANNs are a more appropriate method for predicting the behavior of Indices in Capital Markets than the traditional methods of time series analysis. For this purpose we developed an quantitative study, from financial economic indices, and developed two models of RNA-type feedfoward supervised learning, whose structures consisted of 20 data in the input layer, 90 neurons in one hidden layer and one given as the output layer (Ibovespa). These models used backpropagation, an input activation function based on the tangent sigmoid and a linear output function. Since the aim of analyzing the adherence of the Method of Artificial Neural Networks to carry out predictions of the Ibovespa, we chose to perform this analysis by comparing results between this and Time Series Predictive Model GARCH, developing a GARCH model (1.1).Once applied both methods (ANN and GARCH) we conducted the results' analysis by comparing the results of the forecast with the historical data and by studying the forecast errors by the MSE, RMSE, MAE, Standard Deviation, the Theil's U and forecasting encompassing tests. It was found that the models developed by means of ANNs had lower MSE, RMSE and MAE than the GARCH (1,1) model and Theil U test indicated that the three models have smaller errors than those of a naïve forecast. Although the ANN based on returns have lower precision indicator values than those of ANN based on prices, the forecast encompassing test rejected the hypothesis that this model is better than that, indicating that the ANN models have a similar level of accuracy . It was concluded that for the data series studied the ANN models show a more appropriate Ibovespa forecasting than the traditional models of time series, represented by the GARCH model

PREDIÇÃO DE LINKS EM REDES SOCIAIS ACADÊMICAS: UMA ANÁLISE DA INFLUÊNCIA DAS CARACTERÍSTICAS DA REDE NA EFICÁCIA DOS ALGORITMOS BASEADOS NA TOPOLOGIA

A predição de links em redes sociais tem sido objeto de estudo em um crescente número de artigos científicos e comerciais devido à grande oferta de bases de dados com representações das relações entre pessoas e também devido à facilidade de acesso a recursos computacionais para análise dessas redes. Prever conexões em redes sociais acadêmicas contribui para o crescimento científico, facilitando a colaboração entre pesquisadores com potencial de contribuição mútua. Este trabalho busca identificar as características das redes levam a uma maior eficiência na predição de links feita por algoritmos baseados na topologia. As características serão isoladas conjuntos de dados e alguns experimentos serão repetidos com diferentes algoritmos para buscar identificar tendências. As análises feitas neste trabalho poderão ajudar a compreender melhor a dinâmica das redes sociais acadêmicas e também contribuir na escolha dos melhores algoritmos de predição de link para cada tipo de rede social.