Figure 2From: Bayesian inference of transition matrices from incomplete graph data with a topological priorIllustration of methods to infer transition matrices from repeated interactions on a downstream task of clustering. The first panel shows the ground truth clusters encoded in the transition probabilities of a geometric random network with Euclidean metric in which we observe small set of interactions. We use three different inference methods to construct a transition matrix and detect communities using InfoMap. The other panels show the clusters detected using the transition matrix inferred with a frequentist, noninformative Bayesian, and our approach - BaCon. BaCon finds clusters that are closer to the ones detected from the ground truth matrixBack to article page