“…This model belongs to a larger family of generative models considered in ecology and network theory (9,31,32), and, more generally, the class of latent space models (34), where an edge points from i to j with probability f(s i − s j ) for some function f. These models typically have complicated posterior distributions with many local optima, requiring Monte Carlo methods [for example, (33)] that do not scale efficiently to large networks. In our case, f(s i − s j ) is a Gaussian centered at 1, and the posterior converges to the multivariate Gaussian Eq.…”