Simulation-based inference for rapid Bayesian calibration of epidemiological models

arXiv·low signal

This arXiv paper (2026-06-25) compares simulation-based inference (SBI) using neural posterior estimation against MCMC for Bayesian calibration of mechanistic epidemiological models, where MCMC becomes expensive for high-dimensional nonlinear systems and repeated near-real-time analyses. SBI is presented as a scalable alternative for rapid parameter estimation. The amortized-inference approach generalizes to other scientific-modeling domains needing fast repeated calibration.