Studies in cortical network models

• Ano de defesa: 2020
• Autor(a) principal: Romaro, Cecília
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Mean-field potential; Metaestabilidade; Metastability behavior; Modelagem do córtex somatossensorial; Phase transition; Potencial de campo médio; Redimensionamento de rede neuronal; Rescaling neuron network; Somatosensory cortex model; Transição de fase
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/59/59135/tde-13082020-095609/

This thesis consists of five chapter-articles, each proposing simple solutions to nontrivial questions and problems in computational neuroscience. The first chapter presents a reimplementation of the Potjans-Diesmann (PD) model of the local cortical microcircuitry, and a rescaling method for the number of neurons in the model that is capable of maintaining both the connection probabilities and the behavior of the network activity even when rescaled to 1 % of original size. The second chapter, based on mean field potential, formally explains the scaling method and presents a new method to correct and compensate for the activity of boundary neurons in networks with spatial extension without introducing toroidal connections and/or oscillations. The third chapter introduces spatial extension to the PD model, solves the boundary problem, and studies the spatial (topographic) resolution of the network activity as a consequence of the structural resolution. The fourth chapter, based on phase transition and meta-stability, innovatively studies the false steady state and activity lifetime in networks that do not receive forced external input to keep them active. The fifth chapter contains a characterization of the primary somatosensory cortex network of the rat in terms of a statistical survey of the parameters. It also presents a model of the somatosensory cortex using the stochastic Galves-Löcherbach (GL) neuron, which was constructed based on the somatotopic parameters raised. At the end of the chapter, a method for replacing deterministic leaky integrate-and-fire neurons by GL neurons in neural network models is presented.