Semi-Amortized Bayesian Inference with Reusable Local Posteriors
Abstract
Amortized Bayesian inference uses simulations to perform Bayesian estimation when likelihoods are expensive or analytically unavailable. However, learning the full joint posterior is challenging in high-dimensional models with complex dependencies. We introduce a semi-amortized method that divides parameters into smaller blocks, and uses conditional independence to identify the information needed to infer each block. We train neural estimators for blocks whose posterior updates are computationally expensive or frequently repeated. Each estimator learns from simulations of the relevant model components to approximate the block's full conditional posterior, given the relevant observations and the values of the variables on which it depends. A Markov chain then alternates direct draws from these estimators with conventional updates of the remaining variables. Each neural update uses the current conditioning values and requires neither likelihood evaluation nor an acceptance-rejection step. Where the same conditional structure recurs, one estimator can be reused across blocks and datasets, with an output dimension determined by the block rather than the full model. We illustrate the approach in high-dimensional and computationally expensive hierarchical and spatiotemporal models.
Related
- Subsampled Pseudo-posteriors for Scalable Bayesian Moment-condition Inference
- Amortized Bayesian Inference on Multilevel Models of Arbitrary Structure
- Mitigating Representation Gaps in Amortized Bayesian Inference with Auxiliary Supervision
- Debiased Prior-Data Fitted Networks for Amortized Bayesian Semi-Parametric Inference
- Variational objectives for amortized Bayesian inference in inverse problems: The role of posterior conditioning
BibTeX
@misc{stevenson2026semiamortized,
title = {{Semi-Amortized Bayesian Inference with Reusable Local Posteriors}},
author = {Niek Stevenson and Paul-Christian Bürkner and Andrew Heathcote and Stefan T. Radev},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.18945},
eprint = {2610.18945},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.18945},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Semi-Amortized Bayesian Inference with Reusable Local Posteriors
AU - Niek Stevenson
AU - Paul-Christian Bürkner
AU - Andrew Heathcote
AU - Stefan T. Radev
PY - 2026
DA - 2026/10/13
PB - arXiv
DO - 10.48550/arxiv.2610.18945
UR - https://arxiv.org/abs/2610.18945
AB - Amortized Bayesian inference uses simulations to perform Bayesian estimation when likelihoods are expensive or analytically unavailable. However, learning the full joint posterior is challenging in high-dimensional models with complex dependencies. We introduce a semi-amortized method that divides parameters into smaller blocks, and uses conditional independence to identify the information needed to infer each block. We train neural estimators for blocks whose posterior updates are computationally expensive or frequently repeated. Each estimator learns from simulations of the relevant model components to approximate the block's full conditional posterior, given the relevant observations and the values of the variables on which it depends. A Markov chain then alternates direct draws from these estimators with conventional updates of the remaining variables. Each neural update uses the current conditioning values and requires neither likelihood evaluation nor an acceptance-rejection step. Where the same conditional structure recurs, one estimator can be reused across blocks and datasets, with an output dimension determined by the block rather than the full model. We illustrate the approach in high-dimensional and computationally expensive hierarchical and spatiotemporal models.
ER -