Figure 7. Maximum output signal (level C in.

Fig. 1) as a function of the number of edges M = 100... 700 (steps of 5) for different values of the recovery probability p. The number of nodes is constant (N = 80). For each value of M, 500 network realizations are considered and one input node selected at random. The output excitations are accumulated over T = 300 steps, so that the variability of shorter transients is smoothed out. For each network, the maximum value of the output signal is determined by running the dynamics for each value of 1/κ between 1 and 50 (step of 1), then the observed maximum is averaged over the 500 network realizations with M edges, normalized so that 100 corresponds to the maximal capacity of the output node at p = 1. The plot shows that the sustained activity saturates at large M (dense graphs) to a value dependent on p, with a theoretical maximum value a_max|_p = ₁ = 100 reached for the deterministic dynamics (p = 1).