Fig. S14.

Detailed explanation of how the generative model works. Panel A shows a toy network, with nodes embedded along the perimeter of the unit circle. Given this network and our generative models, we can ask the following question: If we wish to add a new edge to the network according to the wiring rule: P(u, v) = K(u, v)^γ × D(u, v)^η, where will that edge most likely go? To answer this question, we need to first calculate the distance between all pairs of nodes (panel B), whose elements we raise to the power η = − 1 (panel C). The other component we need is the matrix, K(u, v), which represents the non-geometric component. One possible definition of K(u, v) is the product of node degrees (the deg. prod model). Given this particular definition, we set K(u, v) = k_u × k_v. To generate this matrix, we first calculate k_u for all u (panel D). Then we multiply k_u × k_v for all pairs of nodes, {u, v} (panel E). We next raise K(u, v) to the power γ = 1, which we show in panel F. We perform the element-wise multiplication of K(u, v) by D(u, v)^η (panel G, left). Finally, we have to remove the pairs, {u, v}, for which a connection already exists (the gray cells in panel G, right). The nonzero elements of this matrix give us the relative probabilities of where an edge would get placed given this wiring rule. After placing the edge according to these probabilities, the model would return to panel A and the process would repeat itself.