No description available.

Thesis (Ph.D.)--Boston University, 2012 / PLEASE NOTE: Boston University Libraries did not receive an Authorization To Manage form for this thesis or dissertation. It is therefore not openly accessible, though it may be available by request. If you are the author or principal advisor of this work and would like to request open access for it, please contact us at open-help@bu.edu. Thank you. / Effective navigation depends upon reliable estimates of head direction (HD). Visual, vestibular, and outflow motor signals combine for this purpose in a brain system that includes dorsal tegmental nucleus, lateral mammillary nuclei (LMN), anterior dorsal thalamic nucleus (ADN), and the postsubiculum (PoS). Learning is needed to combine such different cues and to provide reliable estimates of HD. A neural model is developed to explain how these three types of signals combine adaptively within the above brain regions to generate a consistent and reliable HD estimate, in both light and darkness. The model starts with establishing HD cells so that each cell is tuned to a preferred head direction, wherein the firing rate is maximal at the preferred direction and decreases as the head turns from the preferred direction. In the brain, HD cells fire in anticipation of a head rotation. This anticipation is measured by the anticipated time interval (ATI), which is greater in early processing stages of the HD system than at later stages. The ATI is greatest in the LMN at -70 ms, it is reduced in the ADN to -25 ms, and non-existing in the last HD stage, the PoS. In the model, these HD estimates are controlled at the corresponding processing stages by combinations of vestibular and motor signals as they become adaptively calibrated to produce a correct HD estimate. The model also simulates how visual cues anchor HD estimates through adaptive learning when the cue is in the animal's field of view. Such learning gains control over cell firing within minutes. As in the data, distal visual cues are more effective than proximal cues for anchoring the preferred direction. The introduction of novel cues in either a novel or familiar environment is learned and gains control over a cell's preferred direction within minutes. Turning out the lights or removing all familiar cues does not change the cells firing activity, but it may accumulate a drift in the cell's preferred direction. / 2031-01-01

Listeners can rapidly adjust how they localize auditory stimuli when consistently trained with spatially discrepant visual feedback. However, relatively little is known about what auditory processing stages are altered by adaptation or the mechanisms that cause the observed perceptual and behavioral changes. Experiments were conducted to test how spatial adaptation generalizes to novel frequencies and the degree to which perceptual recalibration and cognitive adjustment contribute to spatial adaptation. A neural network model was developed to help explain and predict behavioral results. Adaptation was found to generalize across frequency when both training and reference stimuli were dominated by interaural time differences (ITDs), but not when the training stimuli were dominated by interaural level difference (ILDs) and the reference stimuli were dominated by ITDs. These results suggest that spatial adaptation occurs after ITDs are integrated across frequency, but before ITDs and ILDs are integrated. Both perceptual and cognitive changes were found to contribute to short-term auditory adaptation. However, their relative contributions to adaptation depended on the form of the rearrangement of auditory space. For both a magnification and a rotation of auditory space, at least some of the adaptation comes from perceptual recalibration. However, for a magnification of auditory space, cognitive adjustment contributed less to the observed adaptation than for a rotation of auditory space. A hierarchical, supervised-learning model of short-term spatial perceptual, recalibration was developed. Discrepancies between the perceived and correct locations drive learning by adjusting how auditory inputs map to exocentric locations to reduce error. Learning affects locations near the input location through a spatial kernel with limited extent. Model results fit the observed evolution of localization errors and account for individual differences by adjusting only three model parameters: the internal sensory noise, the width of the spatial learning kernel, and the threshold for detecting an error. Results demonstrate how training helps listeners calibrate spatial auditory perception. This work can help inform the design of hearing aids and hearing-protection devices to ensure that listeners receive sufficient information to localize sounds accurately, despite distortions of auditory cues caused by these devices.

Animal vision systems make use of information about an object's motion, color, and form to detect and identify predators, prey and mates. Neurobiological evidence from the macaque monkey indicates that visual processing is separated into two streams: the magnocellular primarily for motion, and the parvocellular primarily for color and form. Two computational systems are developed using key functional properties of the two postulated physiological streams. Each produces invariant representations that act as input to separate copies of a new learning and recognition architecture, Gaussian ARTMAP with covariance terms (GAC). Finally, perceptual experiments are conducted to explore the ability of the human form/color system to detect and recognize targets in photo-realistic imagery. GAC, the component common to both computational systems, retains the on-line learning capabilities of previous ARTMAP architectures, but uses categories that have a location and orientation in the dimensions of the feature space. This architecture is shown to have lower error rates than Fuzzy ARTMAP and Gaussian ARTMAP for all data sets examined, and is used to cluster motion and spectral parameters. For the motion system, local velocity measures of image features are obtained by the method of Convected Activation Profiles. This method is extended and shown to accurately estimate the velocity normal to rotating and translating lines, or of line ends, points, and curves. These local measures are grouped into neighborhoods, and the collection of motions within a neighborhood is described using orientation-invariant deformation parameters. Multiple parameters obtained by examining maneuvering objects are clus­tered, and motions that are characteristic of specific objects are identified. For the form and color system, multi-spectral measurements are made invariant to some fluctuations of local luminance and atmospheric transmissivity by within-band and across-band shunting networks. The resulting color-processed spectral patterns are clustered to enhance the performance of a machine target detection algorithm. Psychophysicists have examined human target detection capabilities primarily via scenes of polygonal targets and distractors on uniform backgrounds. Techniques are developed and experiments are performed to assess human performance of visual search for a complex object in a cluttered scene.

No description available.

No description available.

Convergent evidence suggests that the dorsal medial prefrontal cortex (dmPFC) makes an important contribution to goal-directed action selection. The dmPFC is also part of a network of brain regions that becomes compromised in old age. It was hypothesized that during decision-making, some process of comparison takes place in the dmPFC between the representation of available actions and associated values, and that this process is changed with aging. These hypotheses were tested in aged and young adult rats performing a novel 3-choice, 2-cue decision task. Neuron and local field potential activity revealed that the dmPFC experienced different states during decision and outcome phases of the task, with increased local inhibition and oscillatory (gamma and theta) activity during cue presentation, and increased excitatory neuron activity (among regular firing neurons) at goal zones. Although excitatory and inhibitory activity appeared anti-correlated over phases of the decision task, cross-correlations and the prominent gamma oscillation revealed that excitation and inhibition were highly correlated on the millisecond scale. This "micro-scale" coupling between excitation and inhibition was altered in aged rats and the observed changes were correlated with changes in decision and movement speeds of the aged animals, suggesting a putative mechanism for age-related behavioral slowing. With respect to decision-making, both aged and young adult rats learned over multiple days to follow the rewarded cue in the 3-choice, 2-cue task. Support for the hypothesis that the dmPFC simultaneously represents alternative actions was not found; however, neuron activity selective for particular goal zones was observed. Interestingly, goal-selective neural activity during the decision period was more likely to take place on error trials, particularly on high-performing sessions and when rats exhibited a preference for a particular feeder. A possible interpretation of these patterns is that goal representations in the dmPFC might have sometimes overruled learned habits, which are likely to be involved in following the correct cue and which are known to be supported by other brain regions. These results describe fundamental properties of network dynamics and neural coding in the dmPFC, and have important implications for the neural basis of processing speed and goal-directed action.

The current experiment aims to integrate the neuropsychological and physiological consequences of rape trauma and physical restraint. Given the preponderance of rape on college campuses, it is important for continued research efforts to provide insight into the impact that this traumatic experience may have on the victim. Moreover, it is expected that an improved understanding of these consequences and mechanisms will provide a foundation for prevention and treatment efforts. Within this context, capacity theory provides a basis for appreciating that extreme stress may alter and/or damage neural systems principally associated with the regulatory control or inhibition over brain regions directly involved in the experiential processing and/or comprehension of the traumatic event. The aim of the present experiment was to explore how the experience of rape trauma may alter or diminish this capacity, resulting in deregulation, heightened reactivity, and sensitivity to decomposition from subsequent exposure to these events. It was hypothesized that individuals with resultant capacity limitations would differ in the regulatory control of cynical hostility or denial and sympathetic advances of the autonomic nervous system. Results demonstrated that women who have experienced rape showed decreased frontal regulatory control capacity compared to women who have not experienced rape as evidenced in sympathetic reactivity (heart rate, electrodermal activity, and systolic blood pressure) to frontal lobe stressors. Results are discussed in terms of the extant neuropsychological literature and the implications of observed differences for women who have experienced rape type trauma. / Doctor of Philosophy / Rape as a trauma type is a serious problem with the potential for severe impact on the lives of victims. Based upon past research that provides evidence for neural changes in specific brain pathways that control automatic bodily responses, the current experiment was designed to looked at how the brains of women who have experienced rape may differ from those of women who have not. By presenting women in the experiment with various external stressors and analyzing the automatic reactions of heart rate, blood pressure, and the electrical potential of the skin, it was demonstrated that women who reported a history of rape had increased difficulty controlling their physiological and emotional reactions to stress. The results support the idea that women who have experienced rape may see and experience the world differently than women who have not. The findings of the study are discussed in terms of the overall implications the observed differences may have on the lives of women who have experienced rape and future directions for improved research and interventions, including assessment and treatment, for rape as a trauma type.

Este trabalho apresenta um sistema neural modular, que processa separadamente informações de contexto espacial e temporal, para a tarefa de reprodução de sequências temporais. Para o desenvolvimento do sistema neural foram considerados redes neurais recorrentes, modelos estocásticos, sistemas neurais modulares e processamento de informações de contexto. Em seguida, foram estudados três modelos com abordagens distintas para aprendizagem de seqüências temporais: uma rede neural parcialmente recorrente, um exemplo de sistema neural modular e um modelo estocástico utilizando a teoria de modelos markovianos escondidos. Com base nos estudos e modelos apresentados, esta pesquisa propõe um sistema formado por dois módulos sucessivos distintos. Uma rede de propagação direta (módulo estimador de contexto espacial) realiza o processamento de contexto espacial identificando a seqüência a ser reproduzida e fornecendo um protótipo do contexto para o segundo módulo. Este é formado por uma rede parcialmente recorrente (módulo de reprodução de sequências temporais) para aprender as informações de contexto temporal e reproduzir em suas saídas a seqüência identificada pelo módulo anterior. Para a finalidade mencionada, este mestrado utiliza a distribuição de Gibbs na saída do módulo para contexto espacial de forma que este forneça probabilidades de contexto espacial, indicando o grau de certeza do módulo e possibilitando a utilização de procedimentos especiais para os casos de dúvida. O sistema neural foi testado em conjuntos contendo trajetórias abertas, fechadas, e com diferentes situações de ambigüidade e complexidade. Duas situações distintas foram avaliadas: (a) capacidade do sistema em reproduzir trajetórias a partir de pontos iniciais treinados; e (b) capacidade de generalização do sistema reproduzindo trajetórias considerando pontos iniciais ou finais em situações não treinadas. A situação (b) é um problema de difícil ) solução em redes neurais devido à falta de contexto temporal, essencial na reprodução de seqüências. Foram realizados experimentos comparando o desempenho do sistema modular proposto com o de uma rede parcialmente recorrente operando sozinha e um sistema modular neural (TOTEM). Os resultados sugerem que o sistema proposto apresentou uma capacidade de generalização significamente melhor, sem que houvesse uma deterioração na capacidade de reproduzir seqüências treinadas. Esses resultados foram obtidos em sistema mais simples que o TOTEM. / This work presents a new modular neural system to deal separately with spatial and temporal context information, during temporal sequence processing. Given the initial and final states of the sequence, the neural system can reproduce the whole sequence linking these points. The proposed model involves concepts on recurrent neural networks, stochastic models, modular neural systems and context information processing. Three models based on distinct approaches to learn temporal sequences were particularly important in this work: a partially recurrent neural network, a modular neural system and a stochastic model based on the Hidden Markov Models theory. This master thesis presents a new modular neural system composed of two supervised neural networks. A feedforward neural network (spatial context estimator) to identify the desired sequence to be reproduced and to provide a spatial context prototype to the second module. This is a partially recurrent neural network to reproduce the sequence identified by the former module. Moreover, the first module employs the Gibbs distribution in the spatial context estimator outputs in such a way to obtain the uncertainty of the sequence identification task. Thus, with these probability values, special procedures may be used whenever a doubt occurs. The proposed system was evaluated in different domains containing open and closed sequences with different levels of complexity due to space dimension and level of ambiguity of the trained trajectories. The system was evaluated according to its ability to reproduce the sequence whenever versions of the initial and final points are provided. A version may be exactly the points seen during the training stage or points trained as intermediate states. The latter is considered a difficult task for recurrent neural networks due to the lack of temporal context information. Experiments were done comparing the performance of the proposed modular neural system with the performance of a recurrent neural network itself and a modular neural system (a model called TOTEM) for sequence reproduction. The results suggest that the proposed modular neural system presented ability to generalize significant1y better that of the recurrent neural network without deteriorating its ability to reproduce sequences starting from trained situations. The neural system may reproduce the results of the TOTEM with a simpler topology.

Este trabalho apresenta um sistema neural modular, que processa separadamente informações de contexto espacial e temporal, para a tarefa de reprodução de sequências temporais. Para o desenvolvimento do sistema neural foram considerados redes neurais recorrentes, modelos estocásticos, sistemas neurais modulares e processamento de informações de contexto. Em seguida, foram estudados três modelos com abordagens distintas para aprendizagem de seqüências temporais: uma rede neural parcialmente recorrente, um exemplo de sistema neural modular e um modelo estocástico utilizando a teoria de modelos markovianos escondidos. Com base nos estudos e modelos apresentados, esta pesquisa propõe um sistema formado por dois módulos sucessivos distintos. Uma rede de propagação direta (módulo estimador de contexto espacial) realiza o processamento de contexto espacial identificando a seqüência a ser reproduzida e fornecendo um protótipo do contexto para o segundo módulo. Este é formado por uma rede parcialmente recorrente (módulo de reprodução de sequências temporais) para aprender as informações de contexto temporal e reproduzir em suas saídas a seqüência identificada pelo módulo anterior. Para a finalidade mencionada, este mestrado utiliza a distribuição de Gibbs na saída do módulo para contexto espacial de forma que este forneça probabilidades de contexto espacial, indicando o grau de certeza do módulo e possibilitando a utilização de procedimentos especiais para os casos de dúvida. O sistema neural foi testado em conjuntos contendo trajetórias abertas, fechadas, e com diferentes situações de ambigüidade e complexidade. Duas situações distintas foram avaliadas: (a) capacidade do sistema em reproduzir trajetórias a partir de pontos iniciais treinados; e (b) capacidade de generalização do sistema reproduzindo trajetórias considerando pontos iniciais ou finais em situações não treinadas. A situação (b) é um problema de difícil ) solução em redes neurais devido à falta de contexto temporal, essencial na reprodução de seqüências. Foram realizados experimentos comparando o desempenho do sistema modular proposto com o de uma rede parcialmente recorrente operando sozinha e um sistema modular neural (TOTEM). Os resultados sugerem que o sistema proposto apresentou uma capacidade de generalização significamente melhor, sem que houvesse uma deterioração na capacidade de reproduzir seqüências treinadas. Esses resultados foram obtidos em sistema mais simples que o TOTEM. / This work presents a new modular neural system to deal separately with spatial and temporal context information, during temporal sequence processing. Given the initial and final states of the sequence, the neural system can reproduce the whole sequence linking these points. The proposed model involves concepts on recurrent neural networks, stochastic models, modular neural systems and context information processing. Three models based on distinct approaches to learn temporal sequences were particularly important in this work: a partially recurrent neural network, a modular neural system and a stochastic model based on the Hidden Markov Models theory. This master thesis presents a new modular neural system composed of two supervised neural networks. A feedforward neural network (spatial context estimator) to identify the desired sequence to be reproduced and to provide a spatial context prototype to the second module. This is a partially recurrent neural network to reproduce the sequence identified by the former module. Moreover, the first module employs the Gibbs distribution in the spatial context estimator outputs in such a way to obtain the uncertainty of the sequence identification task. Thus, with these probability values, special procedures may be used whenever a doubt occurs. The proposed system was evaluated in different domains containing open and closed sequences with different levels of complexity due to space dimension and level of ambiguity of the trained trajectories. The system was evaluated according to its ability to reproduce the sequence whenever versions of the initial and final points are provided. A version may be exactly the points seen during the training stage or points trained as intermediate states. The latter is considered a difficult task for recurrent neural networks due to the lack of temporal context information. Experiments were done comparing the performance of the proposed modular neural system with the performance of a recurrent neural network itself and a modular neural system (a model called TOTEM) for sequence reproduction. The results suggest that the proposed modular neural system presented ability to generalize significant1y better that of the recurrent neural network without deteriorating its ability to reproduce sequences starting from trained situations. The neural system may reproduce the results of the TOTEM with a simpler topology.