Fig. 7
From: Inferring and validating mechanistic models of neural microcircuits based on spike-train data
Estimation results for neuronal adaptation using synthetic and in vitro data. a Membrane voltage with indicated spike times (top) and adaptation variable (center) of an adaptive leaky I&F neuron in response to a step of mean input (bottom, for small input noise intensity σ). b Example membrane voltage and true adaptation current time series as well as the reconstruction using estimated adaptation parameters Δw, τw (based on 1000 spikes) and the observed spike times. The adaptation current is considered as normalized by the (unknown) membrane capacitance and therefore in units of mV/ms. Method 1 was used for estimation. c Mean and central 50% (i.e., 25–75th percentile) of relative errors between estimated and true values for Δw and τw as a function of number of spikes K. Insets: empirical density of estimated parameter values with true values indicated for K = 200 and K = 800. d Example recorded membrane voltage in response to an injected noisy step current showing spike rate adaptation. e AIC difference (ΔAIC) between the nonadaptive and adaptive leaky I&F models for all seven PYRs and six INTs. f Estimated mean input μ as a function of empirical mean input μI for the adaptive and nonadaptive models (magenta and black symbols, respectively) for two example cells. ϱ denotes Pearson correlation coefficient. g ϱ for input mean and standard deviation for the two models and all neurons. Estimation results in (e–g) were obtained using 15 s long stimuli (corresponding to three repetitions of 5 s stimuli with equal statistics)
