Appears in collection : 2023 - T1B - WS2 - Networks of spiking neurons

We study the dynamics of spontaneous activity patterns in random networks with the structure of the Erdös-Rényi graph. The networks are made of excitatory and inhibitory neurons modeled as nonlinear, two dimensional integrateand-fire neurons with synaptic noise. We localize the different activity patterns on the parameter diagram spanned by the relative inhibitory synaptic strength and the synaptic noise intensity. In the noiseless setup, the networks display transient activity, either asynchronous and non-oscillatory or oscillatory. For weak noise, the activity patterns are asynchronous and non-oscillatory independently of the synaptic strengths. For stronger noise, the activity patterns have oscillatory and synchrony characteristics which depend on the relative inhibitory synaptic strength. In the inhibition-dominated region of the parameter space and for moderate noise intensities, the networks display intermittent switches between oscillatory and low activity (quiescent) states. In the oscillatory state the neuronal voltages alternate between hyperpolarized and depolarized values in a way that resembles transitions between Up and Down cortical states, and in the quiescent state the voltages fluctuate around the resting state. Increase in noise intensity favors transitions from the quiescent to the oscillatory state and hinders the reverse transitions. The oscillatory and quiescent patterns and transitions between them are analysed and interpreted by using a phenomenological global description of the network state combined with local descriptions of individual neurons in their single-neuron phase space.