Figure 2.

Ability of each measure to predict the Derrida coefficient of heterogeneous random networks. Each point corresponds to a sampled network, and its color indicates the network’s sensitivity. The curves are obtained by fitting a power-law function of various measures to the Derrida coefficient, $\delta$: $kp(1-p)$ to the $\delta$ (left), the traditional comparison, $k_{e}p(1-p)$ to $\delta$ (center), and $H{k}_e$ to $\delta$ (right), our proposed measures. Plots for two sizes of networks ($N=20$, top; $N=200$, bottom) are presented here. Plots for $N=50$ and $N=100$ are available in the Supplemental Figure S3. Plots combine networks sampled from all considered P and $\gamma$ parameter values; thus each plot depicts $36,000$ networks. An alternate curve fit is also investigated in the Supplemental Figure S2. All additional plots show qualitatively similar results to those captured in the images presented in this figure.