Modeling high-dimensional time series from large scale brain networks

• Ano de defesa: 2020
• Autor(a) principal: Nascimento, Diego Carvalho do
• 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: Dados de alta-dimensionalidade; Dynamic Models; Graphical models; High-dimensional data analysis; Modelos de grafos; Modelos dinâmicos; Multivariate Time Series; Séries Temporais Multivariadas
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/104/104131/tde-23072020-155937/

Neuroscientists have an urge to understand the effective brain connectivity, through the direction/ correlation of the brain areas, using biosignals, although this task demands to consider the spatiotemporal dependence and some computational constraints. Naturally, the use of large Vector Autoregression (VARs) would be appropriated if did not present a high-dimensionality curse, where the number of parameters is vastly representative. Additionally, shrinkage either in the data or parameter spaces is not trivial towards maintaining its interpretation. Therefore, some modifications were discussed, towards the graph-based model and entropy analysis, adopting the Bayesian approach, addressed by the estimate of the human brain connectivity using electroencephalogram (EEG) signals. As a motivation, we used a study case of neurorehabilitation, regarding the manipulation of human verticality, we are using high-definition transcranial direct current stimulation (HD-tDCS) as a non-invasive modulation