Title:Neural Posterior Estimation on Exponential Random Graph Models: Evaluating Bias and Implementation Challenges

Abstract:Exponential random graph models (ERGMs) are flexible probabilistic frameworks to model statistical networks through a variety of network summary statistics. Conventional Bayesian estimation for ERGMs involves iteratively exchanging with an auxiliary variable due to the intractability of ERGMs, however, this approach lacks scalability to large-scale implementations. Neural posterior estimation (NPE) is a recent advancement in simulation-based inference, using a neural network based density estimator to infer the posterior for models with doubly intractable likelihoods for which simulations can be generated. While NPE has been successfully adopted in various fields such as cosmology, little research has investigated its use for ERGMs. Performing NPE on ERGM not only provides a differing angle of resolving estimation for the intractable ERGM likelihoods but also allows more efficient and scalable inference using the amortisation properties of NPE, and therefore, we investigate how NPE can be effectively implemented in ERGMs.
In this study, we present the first systematic implementation of NPE for ERGMs, rigorously evaluating potential biases, interpreting the biases magnitudes, and comparing NPE fittings against conventional Bayesian ERGM fittings. More importantly, our work highlights ERGM-specific areas that may impose particular challenges for the adoption of NPE.