Integration of inputs by cortical neurons provides the basis for the complex information processing performed in the cerebral cortex. Here, we propose a new analytic framework for understanding integration within cortical neuronal receptive fields. Based on the synaptic organization of cortex, we argue that neuronal integration is a systems--level process better studied in terms of local cortical circuitry than at the level of single neurons, and we present a method for constructing self-contained modules which capture (nonlinear) local circuit interactions. In this framework, receptive field elements naturally have dual (rather than the traditional unitary influence since they drive both excitatory and inhibitory cortical neurons. This vector-based analysis, in contrast to scalarsapproaches, greatly simplifies integration by permitting linear summation of inputs from both "classical" and "extraclassical" receptive field regions. We illustrate this by explaining two complex visual cortical phenomena, which are incompatible with scalar notions of neuronal integration.

A visãao é o sentido humano mais complexo e importante para os processos de cognição e de interação de um indivíduo com o mundo. Neurofiisiologistas buscam identificar e compreender como funcionam os mecanismos celulares envolvidos neste processo. O sistema visual recebe os sinais de imagens captadas pelos olhos e, por meio de transformações e processamento diversos, integra esses sinais em representações de objetos internos perceptuais. O projeto Olho Virtual consegue reconstruir, em três dimensões, modelos de olhos utilizando córneas reais ou simuladas em computador, reproduzindo suas propriedades ópticas captando imagens de maneira satisfatória. Este trabalho introduz, no projeto Olho Virtual, um modelo computacional de retina baseado no modelo biológico, capaz de reproduzir as propriedades das células cones e bastonetes em suas distribuições radiais e também em suas funcionalidades em particular. Além dessa, é apresentado uma modelagem para reprodução dos campos receptivos das células ganglionares presentes na retina, gerando sinais de saída nos sistemas parvo e magno. Por fim são feitas simulações de experimentos psicofísicos com propósito de verificar a validade da modelagem proposta / The vision is the human sense more complex and important in cognitive processes and the interaction of an individual with the world. Neurophysiologists seek to identify and understand how the cellular mechanisms involved in this process work. The visual system receives the image signals captured by the eyes and, through several transformations and processing, integrate those signals into internal representations of perceptual objects. The project Virtual Eye can reconstruct three-dimensional models of eye corneas using real or simulated on the computer, playing their optical properties capturing images satisfactorily. This work introduces the Virtual Eye project, a computational model of retina-based biological model, able to reproduce the properties of rod and cone cells in their radial distributions and also in its functionality in particular. Besides this, a model is presented for reproduction of the receptive fields of ganglion cells present in the retina, generating output signals in the parvo and magno systems. Finally, simulations are made with psychophysical experiments in order to verify the accuracy of the proposed model

Development of brain function is instructed by both genetically-determined processes (nature) and environmental stimuli (nurture). The relative importance of nature and nurture is a major question in developmental neurobiology. In this dissertation, I investigated the role of visual experience in the development and plasticity of the visual pathway. Each neuron that receives visual input responds to a specific area of the visual field- their receptive field (RF). Developmental refinement reduces RF size and underlies visual acuity, which is important for survival. By rearing Syrian hamsters (Mesocricetus auratus) in constant darkness (dark rearing, DR) from birth, I investigated the role of visual experience in RF refinement and plasticity. Previous work in this lab has shown that developmental refinement of RFs occurs in the absence of visual experience in the superior colliculus (SC), but that RFs unrefine and thus enlarge in adulthood during chronic DR. Using an in vivo electrophysiological approach, I show that, contrary to a widely held view, visual experience is not necessary for refinement of RFs in primary visual cortex (V1). In both SC and V1, RFs refine by postnatal day (P) 60, but enlarge by P90 with chronic DR. One week of visual experience was sufficient to prevent RF enlargement in SC and V1. How normal sensory experience prevents plasticity in mature circuits is not well understood. Using an in vitro electrophysiological approach, I demonstrated that GABAergic inhibition is reduced in DR SC, which in turn affects short-term (but not long-term) synaptic plasticity. The level of GABABR-mediated short-term synaptic depression (STD) that occurs during high-frequency afferent stimulation, such as occurs during vision, is reduced by DR. Using a computational model of RF size, I propose that, in addition to the effect of reduced inhibition, reduced STD of excitation could contribute to enlarged RFs. This work provides insight into mechanisms of development and plasticity of the nervous system. How plasticity is restricted in mature circuits is of fundamental importance in neuroscience and could instruct therapies to prevent maladaptive plasticity in disease and to enhance recovery of function in adults.

A visãao é o sentido humano mais complexo e importante para os processos de cognição e de interação de um indivíduo com o mundo. Neurofiisiologistas buscam identificar e compreender como funcionam os mecanismos celulares envolvidos neste processo. O sistema visual recebe os sinais de imagens captadas pelos olhos e, por meio de transformações e processamento diversos, integra esses sinais em representações de objetos internos perceptuais. O projeto Olho Virtual consegue reconstruir, em três dimensões, modelos de olhos utilizando córneas reais ou simuladas em computador, reproduzindo suas propriedades ópticas captando imagens de maneira satisfatória. Este trabalho introduz, no projeto Olho Virtual, um modelo computacional de retina baseado no modelo biológico, capaz de reproduzir as propriedades das células cones e bastonetes em suas distribuições radiais e também em suas funcionalidades em particular. Além dessa, é apresentado uma modelagem para reprodução dos campos receptivos das células ganglionares presentes na retina, gerando sinais de saída nos sistemas parvo e magno. Por fim são feitas simulações de experimentos psicofísicos com propósito de verificar a validade da modelagem proposta / The vision is the human sense more complex and important in cognitive processes and the interaction of an individual with the world. Neurophysiologists seek to identify and understand how the cellular mechanisms involved in this process work. The visual system receives the image signals captured by the eyes and, through several transformations and processing, integrate those signals into internal representations of perceptual objects. The project Virtual Eye can reconstruct three-dimensional models of eye corneas using real or simulated on the computer, playing their optical properties capturing images satisfactorily. This work introduces the Virtual Eye project, a computational model of retina-based biological model, able to reproduce the properties of rod and cone cells in their radial distributions and also in its functionality in particular. Besides this, a model is presented for reproduction of the receptive fields of ganglion cells present in the retina, generating output signals in the parvo and magno systems. Finally, simulations are made with psychophysical experiments in order to verify the accuracy of the proposed model

The key to understanding vision is to acquire insight into the sensory coding of indi- vidual neurons. To this end, major advances were done over the past 50 years in fitting models to neural data to identify the mapping from sensory space to neural responses. Especially the advance of DNNs in neuroscience allowed for model fits with excellent predictive power. However, such advanced neural models are complex, and their poor in- terpretability has so far hindered deeper understanding of the principles of visual coding. To address this issue, a recent study proposed a method which identifies the stimulus that activates the neuron the most. However, the sensory coding of highly non-linear neurons, which are abundant already at the earliest stages of visual processing, is too complex for a single stimulus to sufficiently characterize it. A more robust way to char- acterize this coding is through identifying the input sub-space within which the neuron is activated identically - i.e. finding invariances of the neuron's sensory representation. In this thesis, a novel approach for finding such invariant stimuli is proposed. The proposed technique is based on a generator neural network, which maps Gaussian noise from latent space to a stimulus set which equally activates a given neuron. The method demonstrated the...

Cat hair receptors were studied to determine whether they could be distinguished by the following receptive field characteristics: thickness of innervated guard hairs, distance between innervated follicles and receptive field size. Initially the receptors were classified as G1, GI, G2 or D on the basis of their velocity requirements for excitation, their degree of linear directionality, their vibrational sensitivity, and whether they were activated by movement of down hairs. It was found that the thickest guard hairs on the posterior aspect of a cat's hindleg were usually 4-5 times thicker than the thinnest guard hairs from the same area and that G1, GI and G2 neurons innervated the full range of guard hair thicknesses available. Although there was a tendency for thicker guard hairs to be more heavily innervated, none of the neurons studied innervated thick guard hairs exclusively. While movement of the down hair and most guard hairs within D-mechanoreceptive fields easily evoked activity, a few guard hairs were regularly found for which mechanical displacement did not elicit a discharge even though they were well within the receptive field. Receptive field sizes and nearest neighbor distances between innervated follicles were smaller for D than for G1, GI and G2 receptors and greater for G1 than GI and G2 receptors.

Rapid receptive field (RF) reorganization of somatosensory neurons in the rat dorsal horn was examined using extracellular single unit recording. Subcutaneous injection of lidocaine into RFs of dorsal horn neurons results in expansion of their RFs within minutes. The expanded RFs appear adjacent to or/and proximal to original RFs. Out of 63 neurons tested, 36 (58%) show RF reorganization. The data suggest that dorsal horn of spinal cord is one of the initial sites for RF reorganization. The neural mechanisms of this effect are not well understood. We propose that changes in biophysical properties (membrane conductance, length constant) of the neurons resulting from lidocaine injection contribute to RF reorganization. Iontophoretic application of glutamate onto dorsal horn neurons that show lidocaine induced RF's expansion were used to test the model. Application of glutamate produced reduction of reorganized RFs in 9 of 20 (45%) tested cells. Application of NBQX produced no effect on either original or expanded RFs indicate that RF shrinkage effects of glutamate involve NMDA receptors. The results are consistent with the prediction of the proposed model. Subcutaneous injection of capsaicin into tactile RFs of low threshold mechanoreceptive dorsal horn neurons produced no effect on the RF sizes that are consistent with other studies. Following the injection, the original RFs were completely silenced (46%) or remained responsive (54%).

Deafferented cortical visual areas exhibit topographical plasticity such that their constituent neural populations adapt to the loss of sensory input through the expansion and eventual remapping of receptive fields to new regions of space. Such representational plasticity is most compelling in the long-term (months or years) but begins within seconds of retinal deafferentation (short-term plasticity). The neural mechanism proposed to underlie topographical plasticity is one of disinhibition whereby long-range horizontal inputs are "unmasked" by a reduction in local inhibitory drive. In this dissertation, four experiments investigated the neural mechanisms of short-term visual plasticity and disinhibition in humans using a combination of psychophysics and event-related potentials (ERPs). Short-term visual plasticity was induced using a stimulus-induced analog of retinal deafferentation known as an artifical scotoma. Artificial scotomas provide a useful paradigm for the study of short-term plasticity as they induce disinhibition but are temporary and reversible. Experiment 1 measured contrast response functions from within the boundaries of an artificial scotoma and evaluated them relative to a sham control condition. Changes in the contrast response function suggest that disinhibition can be conceived of in terms of two dependent but separable processes: receptive field expansion and unrestricted neural gain. A two-process model of disinhibition is proposed. A complementary ERP study (Experiment 2) recorded visual evoked potentials elicited by probes appearing within the boundaries of an artificial scotoma. Results revealed a neural correlate of disinhibition consistent with origins in striate and extrastriate visual areas. Experiment 3 and 4 were exploratory examinations of the representation of space surrounding an artificial scotoma and revealed a neural correlate of invading activity from normal cortex. Together, the results of these four studies strengthen the understanding of the neural mechanisms that underlie short-term plasticity and provide a conceptual framework for their evaluation.

No description available.

[pt] Esta dissertação propõe a Rede Generativa Adversarial por Convolução Rápida de Fourier (FCGAN). Essa abordagem inovadora utiliza convoluções no domínio da frequência para permitir que a rede opere com um campo receptivo de abrangência global. Devido aos seus campos receptivos pequenos, GANs baseadas em convoluções tradicionais enfrentam dificuldades para capturar padrões estruturais e geométricos. Nosso método utiliza Convoluções Rápidas de Fourier (FFCs), que usam Transformadas de Fourier para operar no domínio espectral, afetando globalmente os canais da imagem. Assim, a FCGAN é capaz de gerar imagens considerando informações de todas as localizações dos mapas de entrada. Essa nova característica da rede pode levar a um desempenho errático e instável. Mostramos que a utilização de normalização espectral e injeções de ruído estabilizam o treinamento adversarial. O uso de convoluções espectrais em redes convolucionais tem sido explorado para tarefas como inpainting e super-resolução de imagens. Este trabalho foca no seu potencial para geração de imagens. Nossos experimentos também sustentam a afirmação que features de Fourier são substitutos de baixo custo operacional para camadas de self-attention, permitindo que a rede aprenda informações globais desde camadas iniciais. Apresentamos resultados qualitativos e quantitativos para demonstrar que a FCGAN proposta obtém resultados comparáveis a abordagens estado-da-arte com profundidade e número de parâmetros semelhantes, alcançando um FID de 18,98 no CIFAR-10 e 38,71 no STL-10 - uma redução de 4,98 e 1,40, respectivamente. Além disso, em maiores dimensões de imagens, o uso de FFCs em vez de self-attention permite batch-sizes com até o dobro do tamanho, e iterações até 26 por cento mais rápidas. / [en] This thesis proposes the Fast Fourier Convolution Generative Adversarial Network (FCGAN). This novel approach employs convolutions in the frequency domain to enable the network to operate with a channel-wide receptive field. Due to small receptive fields, traditional convolution-based GANs struggle to capture structural and geometric patterns. Our method uses Fast Fourier Convolutions (FFCs), which use Fourier Transforms to operate in the spectral domain, affecting the feature input globally. Thus, FCGAN can generate images considering information from all feature locations. This new hallmark of the network can lead to erratic and unstable performance. We show that employing spectral normalization and noise injections stabilizes adversarial training. The use of spectral convolutions in convolutional networks has been explored for tasks such as image inpainting and super-resolution. This work focuses on its potential for image generation. Our experiments further support the claim that Fourier features are lightweight replacements for self-attention, allowing the network to learn global information from early layers. We present qualitative and quantitative results to demonstrate that the proposed FCGAN achieves results comparable to state-of-the-art approaches of similar depth and parameter count, reaching an FID of 18.98 on CIFAR-10 and 38.71 on STL-10 - a reduction of 4.98 and 1.40, respectively. Moreover, in larger image dimensions, using FFCs instead of self-attention allows for batch sizes up to twice as large and iterations up to 26 percent faster.