Figure 2

Prehistory heat maps showing the density of the final N stochastic events in 400 stochastic simulations that resulted in switching for two networks. Trajectories are projected into two network bins, denoted by subscripts b and b′. (a) Network with a power-law degree distribution and parameters βλ = 1.625 and \({f}_{i}=0\forall i\). The the horizontal and vertical axes denote, respectively: the fraction of nodes in state s _i  = 1 with degrees 70 ≤ k _i  ≤ 235 and 10 ≤ k _i  ≤ 12. (b) Corresponding heat map for a Facebook network with parameters βλ = 1.37 and \({f}_{i}=0.02\forall i\). The the horizontal and vertical axes denote, respectively: the fraction of nodes in state s _i  = 1 with eigenvector centrality η _i  ≥ 0.0802 and 0.0383 ≤ η _i  ≤ 0.0580. Network details can be found in SI. 3. Arrows indicate the direction in time.