TBD

Apply modern machine learning techniques, developed for the global analysis of complex dynamical systems, to sleep data in order to understand the transitions and underlying processes in sleep across the night and the extent to which observed transitions map onto the conventional sleep stages. By learning dynamical systems or associated transfer operators from data, it is possible to compute characteristic features and their associated time scales automatically. Furthermore, this global approach will allow us to study how these properties and also the connectivity structure---representing how different brain regions interact---change over time.