📝 SummarXiv

Abstract

Temporal networks consist of timestamped directed interactions rather than static links. These links may appear continuously in time, yet few studies have directly tackled the continuous-time modeling of networks. Here, we introduce a maximum entropy approach to temporal networks and with basic assumptions on constraints, the corresponding network ensembles admit a modular and interpretable representation: a set of global time processes -an inhomogeneous Poisson or a Hawkes process- and a static maximum-entropy (MaxEnt) edge, e.g. node pair, probability. This time-edge labels factorization yields closed-form log-likelihoods, degree/unique-edge expectations, and yields a whole class of effective generative models. We provide maximum-entropy derivation of a log-linear Hawkes/NHPP intensity for temporal networks via functional optimization over path entropy, connecting inhomogeneous Poisson modeling -e.g. Hawkes models- to MaxEnt network ensembles. Global Hawkes time layers consistently improve log-likelihood over generic NHPP, while the MaxEnt edge labels recover strength constraints and reproduce expected unique-degree curves. We discuss the limitations of this unified framework and how it could be integrated with calibrated community/motif tools, Hawkes calibration procedures, and (neural) kernel estimation.