Amortized Bayesian Workflow

TL;DR

This paper introduces an adaptive Bayesian inference workflow that combines fast amortized neural network-based methods with accurate MCMC sampling, guided by diagnostics to optimize speed and precision across many datasets.

Contribution

It presents a novel integrated workflow that dynamically switches between amortized inference and MCMC, improving efficiency and accuracy in large-scale Bayesian analysis.

Findings

Significant efficiency gains on synthetic and real datasets

Maintains high posterior quality with adaptive method switching

Effective on tens of thousands of datasets

Abstract

Bayesian inference often faces a trade-off between computational speed and sampling accuracy. We propose an adaptive workflow that integrates rapid amortized inference with gold-standard MCMC techniques to achieve a favorable combination of both speed and accuracy when performing inference on many observed datasets. Our approach uses principled diagnostics to guide the choice of inference method for each dataset, moving along the Pareto front from fast amortized sampling via generative neural networks to slower but guaranteed-accurate MCMC when needed. By reusing computations across steps, our workflow synergizes amortized and MCMC-based inference. We demonstrate the effectiveness of this integrated approach on several synthetic and real-world problems with tens of thousands of datasets, showing efficiency gains while maintaining high posterior quality.