Fig. 10. Contributions to control kernel size for random networks.

Compare to Fig. 4. For descriptions of random ensembles, see “Methods”. A–C Control kernel contributions for random threshold networks (including biased and balanced cases). These networks often require additional control nodes, but the relative number of additional nodes is still often smaller or comparable to the first-order approximation given by the minimal distinguishing nodes. D–F Control kernel contributions for random truth table networks. These networks are typically easier to control. In D, we also see some networks for which all attractors are cycles and none are controllable using static pinning.