Vision research has uncovered computational principles that generalize across species and brain area. However, these biological mechanisms are not frequently implemented in computer vision algorithms. In this thesis, models suitable for application in computer vision were developed to address the benefits of two biologically-inspired computational principles: multi-scale sampling and active, space-variant, vision. The first model investigated the role of multi-scale sampling in motion integration. It is known that receptive fields of different spatial and temporal scales exist in the visual cortex; however, models addressing how this basic principle is exploited by species are sparse and do not adequately explain the data. The developed model showed that the solution to a classical problem in motion integration, the aperture problem, can be reframed as an emergent property of multi-scale sampling facilitated by fast, parallel, bi-directional connections at different spatial resolutions. Humans and most other mammals actively move their eyes to sample a scene (active vision); moreover, the resolution of detail in this sampling process is not uniform across spatial locations (space-variant). It is known that these eye-movements are not simply guided by image saliency, but are also influenced by factors such as spatial attention, scene layout, and task-relevance. However, it is seldom questioned how previous eye movements shape how one learns and recognizes an object in a continuously-learning system. To explore this question, a model (CogEye) was developed that integrates active, space-variant sampling with eye-movement selection (the where visual stream), and object recognition (the what visual stream). The model hypothesizes that a signal from the recognition system helps the where stream select fixation locations that best disambiguate object identity between competing alternatives. The third study used eye-tracking coupled with an object disambiguation psychophysics experiment to validate the second model, CogEye. While humans outperformed the model in recognition accuracy, when the model used information from the recognition pathway to help select future fixations, it was more similar to human eye movement patterns than when the model relied on image saliency alone. Taken together these results show that computational principles in the mammalian visual system can be used to improve computer vision models.

Made available in DSpace on 2015-03-26T13:23:45Z (GMT). No. of bitstreams: 1 texto completo.pdf: 2926433 bytes, checksum: 1d89c84f869b552ddc31a34c52d1ff66 (MD5) Previous issue date: 2012-06-19 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / The Brazilian poultry industry is the activity that has the largest and most advanced technological collection from the agricultural sector in the country, which does not imply that productive problems still do not persist, specifically related to the ambience. In order to optimize the production performance in poultry sector, it is necessary to adequate the aviaries, with techniques that address the thermal and hygienic issues of the internal environment of the grow out houses and a more efficient use of energy. This study assessed the thermal and air internal environments in three grow out houses of different types, representative of poultry production in South America: the first one set with a positive pressure ventilation system in tunnel mode, the second one set with lateral positive pressure ventilation system and the third with negative pressure ventilation system in tunnel mode. Data have been collected during the first phase of growth, in the winter and spring of 2011. Air quality analyzes have been conducted verifying environment s concentrations of CO, CO2 and NH3 and thermal comfort was assessed, through maps of temperature, relative humidity and ITU, in addition to modeling of the internal thermal behavior in CFD for the grow out house with negative pressure ventilation system in tunnel mode. In each chapter, statistical analyzes have been conducted specific to the thermal environment, air quality, and finally to validate the computational model. / A avicultura industrial brasileira é a atividade que possui o maior e mais avançado acervo tecnológico dentre o setor agropecuário no País, o que não implica que não persistam ainda problemas produtivos, especificamente no relacionado com a ambiência. Com o fim de otimizar o desempenho produtivo no setor avícola, faz-se necessária a adequação do ambiente de criação das aves, com técnicas que atendam as questões térmicas e higiênicas do ambiente interno dos aviários e com maior eficiência energética. Este trabalho avaliou os ambientes térmico e aéreo internos em três galpões de diferentes tipologias representativos da produção avícola da América do Sul: o primeiro com sistema de ventilação de pressão positiva em modo túnel, o segundo com sistema de ventilação de pressão positiva lateral e o terceiro com sistema de ventilação de pressão negativa em modo túnel, durante a primeira fase de crescimento no inverno e na primavera de 2011. Foram feitas análises da qualidade do ar em termos das concentrações de CO, CO2 e NH3 do ambiente e do conforto térmico que incluem mapas de temperatura, umidade relativa do ar e ITU, além da modelagem do comportamento térmico interno em CFD para o galpão com sistema de ventilação de pressão negativa em modo túnel. Em cada capítulo, foram feitas análises estatísticas específicas para o ambiente térmico, a qualidade do ar e, finalmente, para validação do modelo computacional.

Submitted by Mario BC (mario@bc.ufrpe.br) on 2016-06-30T12:04:31Z No. of bitstreams: 1 Adriano Luiz Normandia Monteiro.pdf: 2165253 bytes, checksum: a56a08a2a876133289d03517e9c8d77a (MD5) / Made available in DSpace on 2016-06-30T12:04:31Z (GMT). No. of bitstreams: 1 Adriano Luiz Normandia Monteiro.pdf: 2165253 bytes, checksum: a56a08a2a876133289d03517e9c8d77a (MD5) Previous issue date: 2007-05-31 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / This study has been developed in two locations which are typical of communal agriculture in the Northeast semi-arid, in alluvial valleys: Nossa Senhora do Rosário Farm, municipal district of Pesqueira, in the “agreste” area of Pernambuco State and another in the municipal district of Belo Jardim-PE. The objective was to evaluate a computational model to simulate scenarios involving salinization in alluvial soils in the semi-arid Northeast and to simulate the reduction effect on soil salinity in drainage leaching lisimeters and the precipitation as an element to reduce salt content. Based on numeric simulations with finite elements, using the model HYDRUS-1D, it was evaluated flow and transport of salts. For model calibration, of the model field data of water content and soil water pressure head, electric conductivity of soil solution, in adition to soil water characteristic parameters. Using measured actual evapotranspiration of the crop separated: potential evaporation (Ep) and potential transpiration (Tp), and precipitation, besides irrigation with different leaching depths, it was verified that the model is a satisfactory tool to simulate flow and transport in the studied situations. The leach should be used as handling for reduction of the salinity in the root zone since effective precipitations can drain in the winter periods. Incomplete leaching can be an effective alternative to reduce salinity in the root zone, when rainfall might occur. / O presente estudo foi desenvolvido em duas localidades que são modelos típicos de pequena agricultura familiar no semi-árido do Nordeste, em vales aluviais: a Fazenda Nossa Senhora do Rosário, localizada do município de Pesqueira-PE, região Agreste do Estado de Pernambuco, e em lotes irrigados na zona de rural do município de Belo Jardim, região Agreste do Estado de Pernambuco. O objetivo foi calibrar e validar modelo computacional para simular cenários que envolvam o transporte de sais em solos aluviais no semi-árido nordestino, e prever o efeito de lâminas de lixiviação e da precipitação no controle da salinidade do solo em lisímetros de drenagem. Com base em simulações numéricas, utilizando o modelo HYDRUS-1D, avaliou-se o fluxo e o transporte de sais. Para calibração do modelo foram utilizados dados do potencial matricial da água no solo, condutividade elétrica da solução do solo, além de dados da curva característica de retenção de água no solo. Usando medidas de campo de evapotranspiração potencial, e precipitação, foi possível inferir que o modelo apresentou-se como ferramenta adequada nas simulações realizadas, para diferentes cenários de lâminas de lixiviação. Verificou-se experimentalmente e numericamente que a lixiviação incompleta pode ser utilizada como alternativa de manejo para redução da salinidade na zona radicular, desde que precipitações efetivas possam complementar as lavagens.

Este trabalho representa a continuidade de estudos envolvendo a problemática dos escoamentos com superfície livre, contemplando a análise do fenômeno transiente em canais, a partir do modelo matemático unidimensional de Saint-Venant. Para tanto, é desenvolvido um modelo computacional em linguagem FORTRAN, capaz de avaliar o comportamento do escoamento não permanente. As equações hidrodinâmicas completas são discretizadas por um esquema completamente implícito de diferenças finitas e aplicadas no modelo computacional para a avaliação de dois casos. O modelo é previamente testado para um caso simples, cujos resultados são analisados viabilizando o modelo. No primeiro caso, o modelo é aplicado ao canal de alimentação da Usina Hidrelétrica Monjolinho em São Carlos-SP, para avaliar a necessidade de vertedouro quando se dá o fechamento brusco da turbina, e a ocorrência da entrada de ar na mesma quando da sua abertura repentina. No segundo caso, procurou-se avaliar o desenvolvimento do escoamento no Canal do Trabalhador, responsável pelo abastecimento da cidade de Fortaleza-CE. Com manobras de enchimento e esvaziamento do sistema, é possível determinar o tempo de antecedência de liga-desliga do sistema de recalque a partir das alturas dágua e velocidades de ocorrência, permitindo também a automação para as operações de controle. Em ambos os casos o modelo reproduziu resultados que ilustram com coerência os conceitos pré-estabelecidos, constituindo numa ferramenta útil para análise do fenômeno transiente nos escoamentos em condutos livres. / This work presents a computational model developed in FORTRAN language for the study of unsteady open-channel flows with the use of Saint-Venant one-dimensional equation. The discretization of hydrodynamic equations are presented in a completely implicit method of finite differences and applied in the model for the investigation of two cases, besides the one used previously to test the model. In the first case, the model is applied for a channel that supplies the Monjolinho hydroelectric plant in Sao Carlos SP, aiming to evaluate the need of a spillway when the turbine is closed and the flow abruptly stopped, as well as the occurrence of air entering the turbine when it is opened instantaneously. In the second case, the model simulates the development of the flow in the Trabalhador channel, responsible for the water supply in the city of Fortaleza - CE, in order to make possible the automation of operational control, based on data of flow velocity and water level. In both cases the model is presented as a useful tool for the analysis of unsteady open-channel flows, showing results and coherency with theory.

Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-03-07T19:04:39Z No. of bitstreams: 1 lucasdiegomotameneses.pdf: 10631997 bytes, checksum: be72daf41245404c708aab9a8c58d9a1 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-03-10T12:23:50Z (GMT) No. of bitstreams: 1 lucasdiegomotameneses.pdf: 10631997 bytes, checksum: be72daf41245404c708aab9a8c58d9a1 (MD5) / Made available in DSpace on 2017-03-10T12:23:50Z (GMT). No. of bitstreams: 1 lucasdiegomotameneses.pdf: 10631997 bytes, checksum: be72daf41245404c708aab9a8c58d9a1 (MD5) Previous issue date: 2016-09-30 / Modelos computacionais de árvores arteriais são utilizados como substratos geométricos em simulações hemodinâmicas. A construção destes modelos é mandatória para adequada representação das redes vasculares periféricas devido à escassez de dados anatômicos destas redes. Os modelos relatados na literatura são classificados em: anatômico, a parâmetro condensado, fractal e otimizados. O crescimento dos modelos fractais e otimizados dependem de uma lei de bifurcação, que controla a relação entre os raios dos vasos envolvidos na bifurcação através de um expoente. Neste trabalho, investiga-se a construção de modelos otimizados inspirados no método CCO (Constrained Constructive Optimization) usando novas abordagens para a escolha do expoente da lei de bifurcação. Estas estratégias são formuladas com funções degrau e sigmoidal dependentes do número de bifurcações proximais. Os dados morfométricos dos modelos são comparados com outros experimentais e teóricos da literatura. Os resultados obtidos comprovam que o expoente de bifurcação influencia nas estruturas geométrica e topológica dos modelos. / Computational models of arterial trees are used as geometric substrates in hemodynamic simulations. The construction of these models is mandatory for appropriated representation of the peripheral vascular networks due to lack of anatomical data of these networks. The models reported in the literature are classified into: anatomical, lumped parameter, fractal and optimized. The growth of the fractal and optimized models depend on a bifurcation law, which controls the relationship between the radii of the vessels involved in the bifurcation through an exponent. This work investigates the construction of optimized models inspired by the CCO (Constructive Constrained Optimization) method using new approaches to the choice of the exponent of the bifurcation law. Theses strategies are formulated as step and sigmoid functions depend on number of proximal bifurcations. Morphometric data from models are compared with other experimental and theoretical data of the literature. The results obtained show that the bifurcation exponent influences the geometrical and topological structures of the models.

Conditionnement à la peur est un paradigme réussi pour comprendre les substrats neuronaux de l’apprentissage et de l’émotion. Dans cette thèse, je présente deux modèles informatiques des structures du cerveau qui sous-tendent l'acquisition de la peur conditionnée. Le première modèle est utilisé pour enquêter sur l’effet des changements de l’inhibition tonique sur le traitement des informations reçues. On confirme que la diminution de l’inhibition tonique d’une population augmente la réactivité du réseau. Ensuite, le modèle est analysé d’une perspective fonctionnelle et des prédictions qui découlent de cette proposition sont discutées. En outre, je présenterai un modèle systématique, basé sur un type de modèle de conditionnement récemment introduit utilisant des variables latentes. Je propose que l’interaction entre les neurones dans l’amygdale basale code pour l’interface entre ces variables latentes. Le modèle couvre une large gamme d’effets et l’analyse produit un certain nombre de prédictions vérifiables. / Fear conditioning is a successful paradigm for studying neural substrates of emotional learning. In this thesis, two computational models of the underlying neural circuitry are presented. First, the effects of changes in neuronal membrane conductance on input processing are analyzed in a biologically realistic model. We show that changes in tonic inhibitory conductance increase the responsiveness of the network to inputs. Then, the model is analyzed from a functional perspective and predictions that follow from this proposition are discussed. Next, a systems level model is presented based on a recent high-level approach to conditioning. It is proposed that the interaction between fear and extinction neurons in the basal amygdala is a neural substrate of the switching between latent states, allowing the animal to infer causal structure. Important behavioral and physiological results are reproduced and predictions and questions that follow from the main hypothesis are considered.

Alteration in the tone of pulmonary arteries may lead to disease such as pulmonary hypertension often associated with major cardiac complications. This dysfunction is partly in the pulmonary arterial smooth muscle cells (PASMCs) where the excitation-contraction coupling is modified by ion channel behaviour to increase the contractile force. Mathematical models of systemic smooth muscle cells (SMCs) that incorporate electrophysiological and chemomechanical mechanisms to understand the underlying cellular physiology have been successfully employed. Models of pulmonary arterial smooth muscle cells (PASMCs) are only beginning to emerge. Mathematical model prototyping with available experimental data and model investigation from different parameter values is a time-consuming and complex process. This thesis is concerned with the development and validation of mathematical models of excitation-contraction coupling in three types of PASMCs of the rat species, one homogeneous type originating from the distal pulmonary arteries and two from proximal pulmonary arteries. Some key novel additions from previous vascular SMC models include the distinct modelling of Ca2+ in the subplasmalemmal cytosolic region, incorporation of subunit-specific currents from the K+ channel family and a generic G-protein receptor model able to reproduce complex Ca2+ profiles. The main pulmonary and systemic arteries statistically differ in its response to phenylephrine in a wire myograph. The ionic currents of the models were validated against experimental data largely from rat species. The models replicate the recordings of Ca2+ and the resting potential (Em) profiles arising from agonist-induced cytosolic Ca2+ ([Ca2+]i) stimulation (G-protein activation), nifedipine, ryanodine, caffeine and niflumic acid. The distal PASMC model was sensitive to an increase in [Ca2+]i from G-protein activation although were less likely to reproduce Ca2+ oscillations than proximal PASMCs. The proximal models determined the likely proximal PASMC type in literature experiments recording [Ca2+]i and Em. I have developed software that enables other users to simulate Ca2+ and Em changes in SMC studies and the ability to parse a master file describing the mathematical model into different language formats to increase productivity. These models provide a foundation for further studies to better understand PASMC function in the context of normal physiology as well as pathological conditions.

Theories of Extended Mind have evolved in waves to reach the present state of disagreement with regard to whether or not external artefacts become part of the mind when used for memory purposes. A four-step approach has been used to address and resolve this disagreement. First, a new component for models of mind which provides a saliency function is provided. This saliency function corresponds to computational elements found necessary in large mainframe computer systems for handling rich data environments. Second, there is introduced a Computational Model of Memory containing the new component which models the operation of human memory. The Computational Model of Memory contains four interoperative elements including the new component, short-term memory, long-term memory, and a cross-reference associator. Third, the work of Marcin Milkowski is drawn upon to obtain a general method of assessing a computational model’s well-formedness, and the method is applied to prove the adequacy of the Computational Model of Memory. According to Milkowski’s schema, the model satisfies most criteria for a well-formed computational model, including in particular a separation between conceptual elements of the model, and constitutive elements of the model, which while explicitly related, are required to subsist at separate logical conceptual levels. Fourth, the Computational Model of Memory is applied to outstanding arguments in Extended Mind to clarify and resolve several of these arguments. The model serves to highlight where the nature of the disagreement depends upon a category error of reference, and further resolves a key disagreement by demonstrating that the mind may treat external artefacts as an alternative realizable constitutive element of short-term and long-term memory.

Iron is essential for virtually all organisms, yet it can be highly toxic if not properly regulated. Only the Lyme disease pathogen Borrelia burgdorferi has evolved to not require iron (Aguirre et al., 2013).Recent findings have characterised elements of the iron metabolism network, but understanding of systemic iron regulation remains poor. To improve understanding and provide a tool for in silico experimentation, a computational model of human iron metabolism has been constructed. COPASI was utilised to construct a model that included detailed modelling of iron metabolism in liver and intestinal cells. Inter-cellular interactions and dietary iron absorption were included to create a systemic computational model. Parameterisation was performed using a wide variety of literature data. Validation of the model was performed using published experimental and clinical findings, and the model was found to recreate quantitatively and accurately many results. Analysis of sensitivities in the model showed that, despite enterocytes being the only route of iron uptake, almost all control over the system is provided by reactions in the liver. Metabolic control analysis identified key regulatory factors and potential therapeutic targets. A virtual haemochromatosis patient was created and compared to a simulation of a healthy human. The redistribution of control in haemochromatosis was analysed in order to improve our understanding of the condition and identify promising therapeutic targets. Cellular prion protein (PrP) is an enigmatic protein, implicated in disease when misfolded, but its physiological role remains a mystery. PrP was recently found to have ferric-reductase capacity. Potential sites of ferric reduction were simulated and the findings compared to PrP knockout mice experiments. I propose that the physiological role of PrP is in the chemical reduction of endocytosed ferric iron to its ferrous form following transferrin receptor-mediated uptake.

This doctoral thesis is devoted to investigate the problem of establishing connections between Domain Theory and the theory of fuzzy metric spaces, in the sense of Kramosil and Michalek, by means of the notion of a formal ball, and then constructing topological and computational models for (complete) fuzzy metric spaces. The antecedents of this research are mainly the well-known articles of A. Edalat and R. Heckmann [A computational model for metric spaces, Theoret- ical Computer Science 193 (1998), 53-73], and R. Heckmann [Approximation of metric spaces by partial metric spaces, Applied Categorical Structures 7 (1999), 71-83], where the authors obtained nice and direct links between Do- main Theory and the theory of metric spaces - two crucial tools in the study of denotational semantics - by using formal balls. Since every metric induces a fuzzy metric (the so-called standard fuzzy metric), the problem of extending Edalat and Heckmann's works to the fuzzy framework arises in a natural way. In our study we essentially propose two di erent approaches. For the rst one, valid for those fuzzy metric spaces whose continuous t-norm is the minimum, we introduce a new notion of fuzzy metric completeness (the so-called standard completeness) that allows us to construct a (topological) model that includes the classical theory as a special case. The second one, valid for those fuzzy metric spaces whose continuous t-norm is greater or equal than the Lukasiewicz t-norm, allows us to construct, among other satisfactory results, a fuzzy quasi-metric on the continuous domain of formal balls whose restriction to the set of maximal elements is isometric to the given fuzzy metric. Thus we obtain a computational model for complete fuzzy metric spaces. We also prove some new xed point theorems in complete fuzzy metric spaces with versions to the intuitionistic case and the ordered case, respec- tively. Finally, we discuss the problem of extending the obtained results to the asymmetric framework. / Ricarte Moreno, L. (2013). Topological and Computational Models for Fuzzy Metric Spaces via Domain Theory [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/34670 / TESIS