1 volume

No description available.

Thesis (M. A.)--University of North Texas, May, 2007. / Title from title page display. Includes bibliographical references.

No description available.

Os aspectos morfológicos dos neurônios e estruturas neurais, embora potencialmente importantes, têm recebido relativamente pouca atenção na literatura em neurociência. Este trabalho consiste numa substancial parte de um projeto em desenvolvimento no Grupo de Pesquisa em Visão Cibernética voltado para o estudo da relação formal/função neural. Mais especificamente, o presente trabalho dedica particular atenção para a síntese, modelagem e simulação de estruturas neurais morfologicamente realísticas. A tese se inicia com revisões bibliográficas sobre visão biológica e neurociência, direcionadas aos assuntos a serem aqui abordados. Começamos a descrição dos desenvolvimentos com um levantamento, avaliação e proposta de medidas neuromorfométricas adequadas para exprimir as propriedades mais representativas para nosso trabalho, tais como cobertura espacial, complexidade e decaimento eletrônico. Incluímos nessa parte a metodologia utilizada para a geração de neurônios artificiais bidimensionais estatisticamente semelhantes aos naturais. Apresenta-se também a extensão desta metodologia para o caso tridimensional, validada pela análise neuroinorfométrica dos neurônios gerados. Na seqüência, descrevemos o processo de geração de estruturas neurais compostas de neurônios. Considerando modelos com uma camada neural para a codificação de especificidade de orientação, mas sem levar em conta a forma neural, vários casos são simulados, utilizando gradientes na distribuição dos pesos sinápticos e distribuições regulares ou aleatórias (uniformes) dos neurônios na estrutura. A extensão dessas simulações utilizando estruturas que consideram mais detalhadamente a forma neural, usando agora neurônios artificiais gerados pelo método descrito nesta monografia, é apresentada na seqüência. Entre outros efeitos, mostramos que a extensão da arborização dendrítica é um fator determinante da taxa de convergência e seletividade nos modelos, e que gradientes na extensão das arborizações sinápticas são essenciais para a adequada codificação de orientações em módulos cêntricos contendo somatas aleatoriamente distribuídos. / The morphological aspects of neurons and neural structures, although potentially important, have received relatively little attention in the literature in neuroscience. This work consists in a substantial part of a project in development at the Cybernetic Vision Research Group, directed to the study of the form/function relationship. More specifically, the present work dedicates particular attention to the synthesis, modeling, and simulation of morphologically realistic neural structures. The thesis begins with a bibliographic review about biological vision and neuroscience, focusing on the subjects to be here considered. We start the description of the developments with the revision; evaluation and proposal of neuromorphometric measures adequate express the properties more representative to the work, such as spatial cover, complexity and electrotonic decay. We include in this part the methodology used for the generation of bidimensional artificial neurons statistically similar to natural ones. The extension of these developments to the tridimensional case, including their respective validation (performed in terms of neuromorphometric analysis of the generated neurons) is also presented. Next, we describe the generation process of neural structures composed of neurons. Using one-layer neural models for orientation specificity encoding, but without considering the neural shape, several cases are simulated, using gradients in the distribution of the synaptic weights and regular or random (uniform) distributions of the neurons in the structures. The extension of these simulations using structures that consider the neural form in more detail, composed of artificial neurons generated by the described method in this monograph is presented in the sequence. We show that the extension of the dendritic arborization is a determinant factor on the convergence rate and selectivity in the models, and that gradients in the extension of the synaptic arborizations are essentials for the adequate codification of orientations in centric models containing distributed random somata.

Os aspectos morfológicos dos neurônios e estruturas neurais, embora potencialmente importantes, têm recebido relativamente pouca atenção na literatura em neurociência. Este trabalho consiste numa substancial parte de um projeto em desenvolvimento no Grupo de Pesquisa em Visão Cibernética voltado para o estudo da relação formal/função neural. Mais especificamente, o presente trabalho dedica particular atenção para a síntese, modelagem e simulação de estruturas neurais morfologicamente realísticas. A tese se inicia com revisões bibliográficas sobre visão biológica e neurociência, direcionadas aos assuntos a serem aqui abordados. Começamos a descrição dos desenvolvimentos com um levantamento, avaliação e proposta de medidas neuromorfométricas adequadas para exprimir as propriedades mais representativas para nosso trabalho, tais como cobertura espacial, complexidade e decaimento eletrônico. Incluímos nessa parte a metodologia utilizada para a geração de neurônios artificiais bidimensionais estatisticamente semelhantes aos naturais. Apresenta-se também a extensão desta metodologia para o caso tridimensional, validada pela análise neuroinorfométrica dos neurônios gerados. Na seqüência, descrevemos o processo de geração de estruturas neurais compostas de neurônios. Considerando modelos com uma camada neural para a codificação de especificidade de orientação, mas sem levar em conta a forma neural, vários casos são simulados, utilizando gradientes na distribuição dos pesos sinápticos e distribuições regulares ou aleatórias (uniformes) dos neurônios na estrutura. A extensão dessas simulações utilizando estruturas que consideram mais detalhadamente a forma neural, usando agora neurônios artificiais gerados pelo método descrito nesta monografia, é apresentada na seqüência. Entre outros efeitos, mostramos que a extensão da arborização dendrítica é um fator determinante da taxa de convergência e seletividade nos modelos, e que gradientes na extensão das arborizações sinápticas são essenciais para a adequada codificação de orientações em módulos cêntricos contendo somatas aleatoriamente distribuídos. / The morphological aspects of neurons and neural structures, although potentially important, have received relatively little attention in the literature in neuroscience. This work consists in a substantial part of a project in development at the Cybernetic Vision Research Group, directed to the study of the form/function relationship. More specifically, the present work dedicates particular attention to the synthesis, modeling, and simulation of morphologically realistic neural structures. The thesis begins with a bibliographic review about biological vision and neuroscience, focusing on the subjects to be here considered. We start the description of the developments with the revision; evaluation and proposal of neuromorphometric measures adequate express the properties more representative to the work, such as spatial cover, complexity and electrotonic decay. We include in this part the methodology used for the generation of bidimensional artificial neurons statistically similar to natural ones. The extension of these developments to the tridimensional case, including their respective validation (performed in terms of neuromorphometric analysis of the generated neurons) is also presented. Next, we describe the generation process of neural structures composed of neurons. Using one-layer neural models for orientation specificity encoding, but without considering the neural shape, several cases are simulated, using gradients in the distribution of the synaptic weights and regular or random (uniform) distributions of the neurons in the structures. The extension of these simulations using structures that consider the neural form in more detail, composed of artificial neurons generated by the described method in this monograph is presented in the sequence. We show that the extension of the dendritic arborization is a determinant factor on the convergence rate and selectivity in the models, and that gradients in the extension of the synaptic arborizations are essentials for the adequate codification of orientations in centric models containing distributed random somata.

Understanding the mapping between stimulus, behavior, and neural responses is vital for understanding sensory, motor, and general neural processing. We can examine this relationship through the complementary methods of encoding (predicting neural responses given the stimulus) and decoding (reconstructing the stimulus given the neural responses). The work presented in this thesis proposes, evaluates, and analyzes several nonlinear approaches for encoding and decoding that leverage recent advances in machine learning to achieve better accuracy. We first present and analyze a recurrent neural network encoding model to predict retinal ganglion cell responses to natural scenes, followed by a decoding approach that uses neural networks for approximate Bayesian decoding of natural images from these retinal cells. Finally, we present a probabilistic framework to distill behavioral videos into useful low-dimensional variables and to decode this behavior from neural activity.

The work described in this thesis is based on the results of a clinical trial conducted by the research team at the Medical Informatics Unit of the University of Cambridge, which show that the Ballistocardiogram (BCG) has prognostic value in detecting impaired left ventricular function before it becomes clinically overt as myocardial infarction leading to sudden death. The objective of this study is to develop and demonstrate a framework for realising an on-line BCG signal classification model in a portable device that would have the potential to find pathological signs as early as possible for home health care. Two new on-line automatic BeG classification models for time domain BeG classification are proposed. Both systems are based on a two stage process: input feature extraction followed by a neural classifier. One system uses a principal component analysis neural network, and the other a discrete wavelet transform, to reduce the input dimensionality. Results of the classification, dimensionality reduction, and comparison are presented. It is indicated that the combined wavelet transform and MLP system has a more reliable performance than the combined neural networks system, in situations where the data available to determine the network parameters is limited. Moreover, the wavelet transfonn requires no prior knowledge of the statistical distribution of data samples and the computation complexity and training time are reduced. Overall, a methodology for realising an automatic BeG classification system for a portable instrument is presented. A fully paralJel neural network design for a low cost platform using field programmable gate arrays (Xilinx's XC4000 series) is explored. This addresses the potential speed requirements in the biomedical signal processing field. It also demonstrates a flexible hardware design approach so that an instrument's parameters can be updated as data expands with time. To reduce the hardware design complexity and to increase the system performance, a hybrid learning algorithm using random optimisation and the backpropagation rule is developed to achieve an efficient weight update mechanism in low weight precision learning. The simulation results show that the hybrid learning algorithm is effective in solving the network paralysis problem and the convergence is much faster than by the standard backpropagation rule. The hidden and output layer nodes have been mapped on Xilinx FPGAs with automatic placement and routing tools. The static time analysis results suggests that the proposed network implementation could generate 2.7 billion connections per second performance.

Artificial neural networks are computational paradigms which are inspired by biological neural networks (the human brain). Recurrent neural networks (RNNs) are characterized by neuron connections which include feedback paths. This dissertation uses the dynamics of RNN architectures for solving linear and certain nonlinear equations. Neural network with linear dynamics (variants of the well-known Hopfield network) are used to solve systems of linear equations, where the network structure is adapted to match properties of the linear system in question. Nonlinear equations inturn are solved using the dynamics of nonlinear RNNs, which are based on feedforward multilayer perceptrons. Neural networks are well-suited for implementation on special parallel hardware, due to their intrinsic parallelism. The RNNs developed here are implemented on a neural network processor (NNP) designed specifically for fast neural type processing, and are applied to the inverse kinematics problem in robotics, demonstrating their superior performance over alternative approaches.

Thesis (Ph. D.)--Rutgers University, 2009. / "Graduate Program in Biology." Includes bibliographical references (p. 107-114).