Time Series Analysis of Coupled Slow–Fast Neuron Models: From Hurst Exponent to Granger Causality

Abstract

We perform time series analysis of small networks where every node is the slow–fast version of the denatured Morris–Lecar neuron proposed by Schaeffer and Cain. We choose popular coupling strategies from the literature and provide a detailed account of how varying their strength drives the dynamics of the small networks. Algorithms for time series analysis range from measuring their persistence (ability to remember past values), irregularity, chaos, and quasiperiodicity, to synchronization between time nodes within a network. Chaos is observed for inhibitory coupling strengths and for temperatures higher than a reference temperature when the coupling is thermally sensitive. We observe quasi-periodicity when the coupling is very weak and synchronized bursting for high excitatory coupling strength. In certain cases, we also observe decay oscillations. Finally, a causality test is performed to detect whether the dynamics of one neuron influences the dynamics of the other in the coupled system.