Fig. 3 | Nature Communications

Fig. 3

From: Inferring collective dynamical states from widely unobserved systems

Fig. 3

Animal spiking activity in vivo. In neuroscience, m denotes the mean number of spikes triggered by one spike. We estimated \(\hat m\) from spiking activity recorded in vivo in monkey prefrontal cortex, cat visual cortex, and rat hippocampus. a Raster spike plot and population rate at of 50 single units illustrated for cat visual cortex. b MR estimation based on the exponential decay of the autocorrelation r k of a t . Inset: Comparison of conventional and MR estimation results for single units (medians \(\hat m_{\mathrm{C}}\) = 0.057 and \(\hat m\) = 0.954, respectively). c \(\hat m\) estimated from further subsampled cat recordings, estimated with the conventional and MR estimator. Error bars indicate variability over 50 randomly subsampled n out of the recorded 50 channels. d Avalanche size distributions for cat visual cortex (blue) and the networks with AI, reverberating and near-critical dynamics in f. e For all simulations, MR estimation returned the correct distance to instability (criticality) \(\epsilon\) = 1 − m (Supplementary Note 8). In vivo spike recordings from rat, cat, and monkey, clearly differed from critical (\(\epsilon\) = 0) and AI (\(\epsilon\) = 1) states (median \(\hat m\) = 0.98, error bars: 16 to 84% confidence intervals, note that some confidence intervals are too small to be resolved). Opaque symbols indicate that MR estimation was rejected (Supplementary Fig. 5, Supplementary Note 5). Green, red, and yellow arrows indicate \(\epsilon\) for the dynamic states shown in f. f Population activity and raster plots for AI activity, reverberating, and near critical networks. All three networks match the recording from cat visual cortex with respect to number of recorded neurons and mean firing rate