Community detection and reciprocity in networks by jointly modelling pairs of edges

To unravel the driving patterns of networks, most popular models rely on community detection algorithms. However, these approaches are generally unable to reproduce structural features network. Therefore, attempts always made develop that incorporate network properties beside structure. In this work, we present a probabilistic generative model and an efficient algorithm both perform capture reciprocity in networks. Our approach jointly pairs edges with exact 2-edge joint distributions. addition, it provides closed-form analytical expressions for marginal conditional We validate our synthetic data recovering communities, edge prediction tasks, generating networks replicate values observed real also highlight findings two datasets relevant social scientists behavioral ecologists. method overcomes limitations standard algorithms recent through pseudo-likelihood approximation. The inference parameters is implemented by scalable expectation-maximization algorithm, as exploits sparsity dataset. provide open-source implementation code online.