Neste trabalho realizamos um amplo estudo de simulação com o objetivo principal de avaliar o desempenho de carteiras de mínima variância global construídas com base em modelos de previsão da volatilidade que utilizam dados de alta frequência (em comparação a dados diários). O estudo é baseado em um abrangente conjunto de dados financeiros, compreendendo 41 ações listadas na BM&FBOVESPA entre 2009 e 2017. Nós avaliamos modelos de previsão de volatilidade que são inspirados na literatura ARCH, mas que também incluem medidas realizadas. Eles são os modelos GARCH-X, HEAVY e Realized GARCH. Seu desempenho é comparado com o de carteiras construídas com base na matriz de covariância amostral, métodos de encolhimento e DCC-GARCH, bem como com a carteira igualmente ponderada e o índice Ibovespa. Uma vez que a natureza do trabalho é multivariada, e a fim de possibilitar a estimação de matrizes de covariância de grandes dimensões, recorremos à especificação DCC. Utilizamos três frequências de rebalanceamento (diária, semanal e mensal) e quatro conjuntos diferentes de restrições sobre os pesos das carteiras. A avaliação de desempenho baseia-se em medidas econômicas tais como retornos anualizados, volatilidade anualizada, razão de Sharpe, máximo drawdown, Valor em Risco, Valor em Risco condicional e turnover. Como conclusão, para o nosso conjunto de dados o uso de retornos intradiários (amostrados a cada 5 e 10 minutos) não melhora o desempenho das carteiras de mínima variância global. / In this work we perform an extensive backtesting study targeting as a main goal to assess the performance of global minimum variance (GMV) portfolios built on volatility forecasting models that make use of high frequency (compared to daily) data. The study is based on a broad intradaily financial dataset comprising 41 assets listed on the BM&FBOVESPA from 2009 to 2017. We evaluate volatility forecasting models that are inspired by the ARCH literature, but also include realized measures. They are the GARCH-X, the High-Frequency Based Volatility (HEAVY) and the Realized GARCH models. Their perfomances are benchmarked against portfolios built on the sample covariance matrix, covariance matrix shrinkage methods, DCC-GARCH as well as the naive (equally weighted) portfolio and the Ibovespa index. Since the nature of this work is multivariate and in order to make possible the estimation of large covariance matrices, we resort to the Dynamic Conditional Correlation (DCC) specification. We use three different rebalancing schemes (daily, weekly and monthly) and four different sets of constraints on portfolio weights. The performance assessment relies on economic measures such as annualized portfolio returns, annualized volatility, Sharpe ratio, maximum drawdown, Value at Risk, Expected Shortfall and turnover. We also account for transaction costs. As a conclusion, for our dataset the use of intradaily returns (sampled every 5 and 10 minutes) does not enhance the performance of GMV portfolios.

Para verificar a contribuição da audiometria em altas freqüências (9 a 20 kHz) e das emissões otoacústicas - produto de distorção para a identificação de alterações auditivas em indivíduos com desordem pigmentar do tipo vitiligo, 30 indivíduos portadores de vitiligo foram avaliados audiologicamente, e comparados a um grupo controle. Tanto para as altas freqüências (9 a 20 kHz) como para as emissões otoacústicas foram observadas diferenças significantes entre os grupos, que não foram evidenciadas na audiometria convencional (0,25 a 8 kHz). Os resultados sugerem prejuízo nas funções auditivas para o grupo com desordem pigmentar do tipo vitiligo. / In order to verify the contribution of the high frequency audiometry (9 to 20 kHz) and distortion product- otoacoustic emissions for the identification of auditory impairments in individuals with pigmentary disorder type vitiligo, 30 individuals with vitiligo were evaluated, and compared with a control group. As much for the high frequencies (9 the 20 kHz) as for the otoacoustic emissions, significant differences were observed between the groups, that were not evidenced in the conventional audiometry (0,25 the 8 kHz). The results suggest damage in the auditory functions for the group with pigmentary disorder type vitiligo

Electromagnetic (EM) heat exchangers are devices that absorb EM radiation and convert its energy to thermal energy for a specific purpose such as to power a turbine. They have recently been of growing interest, yet the field is predominantly studied with thermal resistance network models and is in need of more rigorous continuum modeling. Homogenization has been used in low and high frequency electromagnetics to describe macroscopic behavior of traveling waves. While dielectric material parameters vary with temperature, coupling the energy equation with Maxwell’s equations, little effort has been made toward homogenization techniques that capture the effects of this dependence, which is necessary to accurately model porous medium heat exchangers. Firstly, we have examined the effect the wave-geometry interactions of high-frequency illumination has on a triple-layer laminate, which approximates the unit cell of a homogenization problem. Secondly, we develop an extension to a high-frequency homogenization (HFH) method developed for photonics. The extension is made by developing a three-dimensional vector-valued HFH of Maxwell’s curl-curl equation that includes dielectric loss. It is validated for a one-dimensional geometry where the exact solution to the scattering problem is known by implementing the Transfer Matrix Method. The HFH model produces perturbation approximations to the dispersion curves showing the nonexistence of band gaps and generates low attenuation outside the band gap regions.

Cette étude a été divisée en 2 parties principales. Dans la première partie, nous examinons la relation entre l'activité des sources neuronales et les HFOs observés sur les électrodes intracérébrales. La deuxième partie traite de l'étude des conditions d'observabilité des HFO sur les électrodes du cuir chevelu. Les simulations ont montré que le modèle de champ neuronal proposé est capable de générer des HFOs montrant une forte ressemblance avec les signaux réels dans les deux cas EEG (cuir chevelu) et SEEG (intracérébral). De plus, nous avons pu relier les mécanismes physiopathologiques (GABA dépolarisant, inhibition directe, activité désynchronisée des populations neuronales) aux différentes caractéristiques morphologiques et spectrales des HFOs intracérébrales. Une hypothèse unifiée pour la production des HFOs et des pointes intercritiques est également formulée. Enfin, nous avons réussi à établir les conditions nécessaires sur l'activité temporelle et l'organisation spatiale des sources neuronales pour observer des HFOs sur les électrodes intracérébrales.En ce qui concerne la deuxième partie, la baisse inexpliquée de fréquence dans les HFOs collectées sur les électrodes du cuir chevelu a été abordée. Nous avons constaté que les mécanismes «non oscillatoires» de la génération de HFOs sont à l'origine de la faible fréquence (<200Hz) des HFOs du cuir chevelu et que le rapport signal / bruit (SNR) influe fortement sur la fréquence des oscillations. De plus, nous avons étudié la topographie des HFOs sur les électrodes du cuir chevelu et analysé comment cette topographie est affectée par différents paramètres (étendue spatiale épileptique, SNR, géométrie 3D). Enfin, nous avons montré que les HFOs du cuir chevelu peuvent être utilisés efficacement pour identifier la zone épileptique lorsque le rapport signal sur bruit des signaux enregistrés est suffisamment élevé. Une perspective de ce travail est l'identification non-invasive de la zone épileptique sans la nécessité d'enregistrements intracérébraux pré-chirurgicaux.Pour les deux études (HFO observés sur les électrodes intracérébrales et du cuir chevelu), un logiciel original et convivial a été développé. Ce logiciel a fortement facilité la simulation des signaux dans l'environnement cerveau/électrode virtuel, signaux obtenus en résolvant le problème direct de l’EEG (projection de la contribution électrique des sources neuronales sur les capteurs). / This study was divided into 2 main parts. In the first part, we address the relationship between the activity of neuronal sources and the HFOs observed on intracerebral electrodes. The second part deals with the investigation of observability conditions of HFOs on scalp electrodes. Simulations showed that the proposed neural field model is capable of generating HFOs showing strong resemblance with real signals in both cases EEG (scalp) and SEEG (intracerebral). Moreover, we were able to relate the pathophysiological mechanisms (depolarizing GABA, feedforward inhibition, desynchronized activity of neuronal populations) to different morphological and spectral features of intracerebral HFOs. A unified hypothesis for generation of HFOs and interictal spikes is also formulated. Finally, we managed to establish the necessary conditions about the temporal activity and the spatial organization of neuronal sources and about for HFOs to be observed on intracerebral electrodes. Regarding the second part, the unexplained drop in frequency in the collected HFOs on scalp electrodes was addressed. We found that the “non-oscillatory” mechanisms of the HFO generation is behind the low frequency (<200Hz) in scalp HFOs and that signal to noise ratio (SNR) heavily impacts the frequency of the oscillations. Moreover, we studied the topography of HFOS on scalp electrodes and analyzed how this topography is affected by different parameters (epileptic spatial extent, SNR, 3D geometry). Finally we showed that scalp HFOs can be effectively used to identify the epileptic zone when the SNR of the recorded signals is sufficiently high. A perspective to this work is the non-invasive identification of epileptic zone without the need for presurgical intracerebral recordings. For the purpose of both studies (HFOs observed on intracerebral & scalp electrodes) an original and user-friendly software package was developed. This software strongly facilitated the simulation of signals in the virtual brain/electrode environment obtained by solving the (S)EEG forward problem (projection of the electric contribution of neuronal sources onto electrode contacts).

This thesis investigates the footfall excitation of higher modes of vibration in low-frequency floor structures. This is motivated by the increased number of floors reportedly failing to meet the required occupants comfort level although being designed in accordance with the current state-of-the-art design guidelines. In particular modern, lightweight, and slender floor structures. The contribution to knowledge of this thesis can be summarised as: quantifying the signal energy of measured walking forces within and above the natural frequency cut-off proposed by the current state-of-the-art design guidelines; quantifying the contribution of higher modes of vibration to the overall response of low-frequency floors to human walking; propose measures to judge the response nature of low-frequency floors, these are the relevant change of the point stiffness and the shape of frequency response functions; proposing a frequency-domain approach that enables designers to include higher modes of vibration in the design against human-induced vibration. It was found that the signal energy of walking forces is distributed well beyond the natural frequency cut-off proposed by the current state-of-the-art design guidelines. Also, the contribution of localised, higher, modes of vibration to the overall response of ultra-lightweight floors was significant. Moreover, it was found that higher modes affect the response of floors of various construction types in one way or another. Hence, it was recommended to consider their contribution in the design of floors against human-induced vibration. Also, it was found that the higher the relative change of the point stiffness the more higher modes contribute to the overall response of floors. Finally, the frequency-domain analysis was found less expensive than time-domain analysis and could result in similarly useful information.

In this thesis we focus on modelling correlation between selected stock markets using high-frequency data. We use time-series of returns of following indices: FTSE, DAX PX and S&P, and Gold and Oil commodity futures. In the first part of our empirical work we compute daily realized correlations between returns of subject instruments and discuss the dynamics of it. We then compute unconditional correlations based on average daily realized correlations and using feedforward neural network (FFNN) to assess how well the FFNN approximates realized correlations. We also forecast daily realized correlations of FTSE:DAX and S&P:Oil pairs using heterogeneous autoregressive model (HAR), autoregressive model of order p (AR(p)) and nonlinear autoregressive neural network (NARNET) and compare performance of these models.

A pesquisa tem por finalidade avaliar dois modelos econométricos de mudanças de preços, que podem ser utilizados em estratégias de arbitragem estatística, o probit ordenado e o de decomposição, estimando seus parâmetros em quatro pregões de mini contratos de dólar futuro negociados na bolsa de valores brasileira. O estudo da negociação em alta frequência com a utilização de dados de transação a transação revela informações relativas à microestrutura de mercado que o ferramental mais tradicional não é capaz de desvendar. Uma das razões é que modelos tradicionais trabalham com variações de preço como variáveis contínuas, enquanto que ao considerar as variações de preço uma variável contínua e não uma variável discreta, como nos modelos aqui avaliados. Este trabalho acrescenta à literatura sobre microestrutura de mercado ao aplicar os modelos estudados em um ativo distinto daqueles avaliados nos papers originais, voltados ao exame do mercado de ações. Esta pesquisa concluiu que os modelos probit ordenado e de decomposição podem ser utilizados para previsão de mini contratos de dólar futuro e que o modelo de decomposição apresenta parâmetros mais significantes. Também concluiu-se que, no modelo probit ordenado, as variáveis de volume e time duration não se apresentaram relevantes na determinação do preço desse contrato e que a quantidade de defasagens utilizadas nos parâmetros estimados pode variar dentre os pregões. / The research aims to evaluate two econometric models of price change, which can be used in strategies of statistical arbitrage, the ordered probit model and the decomposition model, estimating its parameters in four trading sessions of mini US dollar futures contracts traded on the Brazilian Stock Exchange. The study of high frequency trading with the use of trade-by-trade price movements reveals information related to the market microstructure that the more traditional econometric tools are not able to solve when considering the price changes as a continuous variable and not a discrete one, like in the models evaluated here. This work adds to the literature on market microstructure by applying the models studied in an asset different from those evaluated in the original papers, aimed at examining the stock market. This research concluded that the ordered probit and decomposition models can be used to predict mini US dollar futures price changes and that the decomposition model presents more significant parameters. It was also concluded that, in the ordered probit model, the volume and time duration variables were not relevant in determining the price of this contract and that the number of lags used to estimate parameters can vary among the trading sessions.

Submitted by Jadi Castro (jadiana.castro@ucb.br) on 2017-02-21T18:01:09Z No. of bitstreams: 1 AdaoVoneTeixeiradeAssuncaoDissertacao2016.pdf: 1069424 bytes, checksum: 2cec9cf848a40c34902559e8d8f0c95c (MD5) / Approved for entry into archive by Kelson Anthony de Menezes (kelson@ucb.br) on 2017-02-21T18:02:23Z (GMT) No. of bitstreams: 1 AdaoVoneTeixeiradeAssuncaoDissertacao2016.pdf: 1069424 bytes, checksum: 2cec9cf848a40c34902559e8d8f0c95c (MD5) / Made available in DSpace on 2017-02-21T18:02:24Z (GMT). No. of bitstreams: 1 AdaoVoneTeixeiradeAssuncaoDissertacao2016.pdf: 1069424 bytes, checksum: 2cec9cf848a40c34902559e8d8f0c95c (MD5) Previous issue date: 2016-03-31 / Aplicamos o Modelo Autoregressivo de Dura????o Condicional (ACD) Mercado de Futuros de Taxa Interbanc??ria Brasileira. A amostra foi constru??da com base em contratos M??s antes da expira????o para replicar a curva de obriga????es de um m??s eo per??odo estudado Vai de julho de 2013 a setembro de 2015. Utilizamos M??xima Verossimilhan??a Estimativa baseada nas distribui????es de probabilidade mais populares na literatura ACD: Exponencial, gama e Weibull e verificou-se que a estimativa baseada na A distribui????o exponencial foi a melhor op????o para modelar os dados. / We applied the basic Autoregressive Conditional Duration Model (ACD) to the Brazilian Interbank Rate Futures Market. The sample was built using contracts in the month prior to expiration to replicate a one month bond curve and the period studied goes from july of 2013 to september of 2015. We used Maximum Likelihood Estimation based on the most popular probability distributions in the ACD literature: exponential, gamma and Weibull and we found that the estimation based on the exponential distributional was the best option to model the data.

Os Sistemas de Distribuição de Energia Elétrica (SDEE) estão constantemente expostos à ocorrência de faltas, o que torna de primordial importância para as concessionárias que estas sejam localizadas com rapidez e precisão. Com isso, a degradação da confiabilidade do sistema e as perdas econômicas são minimizadas. Com este intuito, diversas pesquisas vêm sendo conduzidas nas últimas décadas tendo em vista o desenvolvimento de métodos computacionais para Localização de Faltas (LF) em SDEE. Embora as técnicas disponíveis atualmente sejam capazes de estimar a distância das faltas com relativa exatidão, algumas características intrínsecas aos SDEE ainda impõem limitações à LF em alimentadores radiais. Em geral, o aspecto econômico requer o uso de dados provenientes de um único terminal de medição. Ainda, a quantidade significativa de ramais laterais tipicamente presentes nos SDEE faz com que o problema das múltiplas estimativas da LF seja considerado como a principal limitação à efetividade das técnicas existentes na atualidade. Neste contexto, este trabalho apresenta uma metodologia baseada na análise dos transitórios de alta frequência gerados pelas faltas, a qual é capaz de fornecer uma estimativa única para a LF em SDEE ramificados a partir de medições somente no terminal local. O ramal em falta e a distância da falta em relação ao terminal de medição são determinados através da comparação entre as frequências características identificadas no espectro do transitório, e as frequências teóricas calculadas para os possíveis caminhos de propagação. Uma das principais contribuições deste trabalho consiste na formulação de um modelo detalhado das linhas de distribuição desequilibradas. A formulação proposta é baseada na modificação nas equações de Carson, considerando a dependência da frequência e a inclusão dos efeitos pelicular e da corrente de retorno pela terra para altas frequências. Como resultado, esta abordagem permite uma maior exatidão na determinação dos parâmetros modais que descrevem a propagação de transitórios em SDEE desequilibrados, eliminando as aproximações adotadas pela maioria das técnicas propostas na literatura atual. A técnica proposta foi avaliada considerando simulações de faltas nos alimentadores IEEE 13 e 34 barras através do ATP/EMTP. Os resultados apresentados incluem diversos cenários de faltas, bem como a comparação com um método de LF para SDEE considerado estado da arte atualmente. / Electric power distribution systems (EPDS) are continuously exposed to faults, therefore fast and accurate fault location is of paramount importance for utilities. Thus, degradation of system reliability and economic losses are minimized. In this sense, several studies have been conducted in the last decades aiming to the development of computational methods for Fault Localization (FL) in EPDS. Although the currently available techniques are able to estimate fault distance with relative accuracy, some intrinsic characteristics of EPDS still impose limitations to FL in radial feeders. In general, the economic aspect requires use of data from one-terminal measurements. Also, typical EPDS have a large number of branches, which makes the problem of multiple FL estimates the main limitation to the effectiveness of the existing techniques. In this context, this work presents a methodology based on the analysis of the high frequency transient generated by faults, which is able to provide a unique FL estimate in branched EPDS by using only one-terminal measurements. The faulted branch and the fault distance from the measurement terminal are determined by correlating the characteristic frequencies identified in the transient spectrum and theoretical frequencies calculated for the possible propagation paths. One of the main contributions of this work is the formulation of a detailed model of unbalanced distribution lines. The proposed formulation is based on the modification of Carson’s equations, considering frequency dependence and inclusion of skin effects and the ground current return at high frequencies. As a result, this approach allows a greater accuracy in determining the modal parameters that describe the transients’ propagation in unbalanced EPDS, thus eliminating the approximations adopted by most of the techniques proposed in the current literature. The proposed technique was evaluated considering fault simulations in the IEEE 13 and 34 nodes feeders through the ATP/EMTP. Results presented include several fault scenarios as well as the comparison with a FL method for SDEE currently considered as the state of the art.

The second most common cause of cancer deaths in the developed world is bowel cancer. Improving the ability to detect and classify lesions as early as possible, allows treatment earlier. The work presented in this thesis is structured around the following detailed aims:Development of high frequency, broadband µUS (micro-ultrasound) imaging transducers through optimization of ultra-thinning processes for lithium niobate (LNO) and fabrication of novel ‘mass-spring’ matching layers using carefully controlled vacuum deposition is demonstrated. The effectiveness of this technique was quantified by applying multiple matching layers to 3 mm diameter 45 MHz LNO µUS transducers using carefully controlled vacuum deposition. The bandwidth of single mass-spring layer µUS transducer was measured to be 46% with an insertion loss of 21 dB. The bandwidth and insertion loss of a multiple matching layer µUS transducer was measured to be 59% and 18 dB respectively. The values were compared with an unmatched transducer which had a bandwidth and insertion loss of 28% and 34 dB respectively. All the experimentally measured values were in agreement with unidimensional acoustic model predictions. µUS tools that can detect and measure microscopic changes in precancerous tissue using a mouse small bowel model with an oncogenic mutation was developed. µUS transducer was used to test the hypothesis that the intestinal tissue morphology of WT (wild type) and ApcMin/+ (adenomatous polyposis coli) diverges with progressing age intervals (60, 90 and 120 days) of mice. A high frequency ultrasound scanning system was designed and the experiments were performed ex vivo using a focused 45 MHz, f-# = 2.85, µUS transducer. The data collected by scanning was used to compute the backscatter coefficients (BSC) and acoustic impedance (Z) of WT and ApcMin/+ mice. The 2D and 3D ultrasound images showed that µUS detects polyps < 500 µm in the scan plane. The measured values of BSC and Z showed differences between normal and precancerous tissue. The differences detected in precancerous murine intestine and human tissue using µUS were correlated with high resolution 3D optical imaging. This novel approach may provide a powerful adjunct to screening endoscopy for improved identification and monitoring, allowing earlier treatment of otherwise undetectable lesions.