Abstract
Neural Posterior Estimation methods for simulation-based inference can beill-suited for dealing with posterior distributions obtained by conditioning onmultiple observations, as they may require a large number of simulator calls toyield accurate approximations. Neural Likelihood Estimation methods cannaturally handle multiple observations, but require a separate inference step,which may affect their efficiency and performance. We introduce a new methodfor simulation-based inference that enjoys the benefits of both approaches. Wepropose to model the scores for the posterior distributions induced byindividual observations, and introduce a sampling algorithm that combines thelearned scores to approximately sample from the target efficiently.
Quick Read (beta)
loading the full paper ...