Fig. 3: Multifractal analysis of neuronal spiking dynamics captures variations in network connectivity while being robust to changes in the thalamic input strength.

a A power-law relationship between the q-order fluctuations and the scale indicates a multifractal structure for spiking dynamics of excitatory neurons of the network in Fig. 2. Axes are plotted using a base-2 logarithmic scale. The average q-order Hurst exponent (b) and the multifractal spectrum (c) of the excitatory cells under different lateral connection probabilities and thalamic input strengths. Curves cluster based on connection probability, not thalamic input strength. d Spike counts of excitatory neurons scale linearly with the intensity of thalamic inputs. Spike counts are calculated for the whole simulation with a duration of 500s. In contrast to the large changes in spike counts, the q-order Hurst exponent (e) and multifractal spectrum (f) are largely invariant to the thalamic input strength. Yet, both multifractal metrics reliably capture changes in network topology (lateral connection probabilities). The average q-order Hurst exponent (g) and multifractal spectrum (h) of Erdös-Rényi (ER), Barabási-Albert (BA), and Watt-Strogatz (WS) network (net) model with varying thalamic input strengths.