Figure 7.

Dynamical properties of spontaneous activity in associative networks. A1, Two examples of spike rasters for associative networks parametrized as indicated by the red and green asterisks from A2. Dynamics at low values of rescaled robustness (red) quickly terminates at a quiescent state. A2, Map of the duration of transient dynamics as a function of rescaled robustness and relative memory load. At high levels of rescaled robustness and memory load, associative networks have long-lasting, transient activity. Isocontour line is drawn as a guide to the eye. B1, Distributions of CV in ISIs for the two parameter settings. The average CV value increases with ρ. B2, Map of the average CV of ISI as a function of rescaled robustness and relative memory load. Isocontour line indicates a region of high CV values that are in general agreement with experimental measurements. C, Same for cross-correlation coefficients of neuron spike trains. D, Same for the anticorrelation coefficient of excitatory and inhibitory postsynaptic inputs received by a neuron. The inputs are normalized by the firing threshold. For the selected parameter configurations, excitatory and inhibitory inputs are tightly balanced (large anticorrelation) despite large fluctuations. A2, B2, C2, D2, Maps were generated based on networks of N = 800 neurons by averaging the results over 100 networks and 100 random initial states for each network and parameter setting.