Network methods for analysing time-series from complex interacting systems #

Ramón Nartallo-Kaluarachchi, Alain Goriely, Renaud Lambiotte, Morten Kringelbach

11:30 Tuesday in 3Q68.

Part of the Mathematical medicine session.

Abstract #

Network analysis has long been associated with the study of the brain whilst neuroimaging data presents neuroscientists with the problem of analysing high-dimensional and non-linear time series. At the interface of these ideas, network-based analysis provides a novel way of extracting information from non-linear time-series. A particularly promising technique is the use of the multiplex visibility network which captures key spatiotemporal information in its structure. Motivated by applications in neural dynamics, we construct the multiplex visibility network on time-series generated by both coupled oscillator networks and multivariate Ornstein-Uhlenbeck processes. In particular, we use multiplex network measures to extract bifurcations and quantify levels of turbulence in complex systems and show that the rate of entropy production in encoded in the degree distribution of the network.