Amortized Factor Inference Networks for Posterior Inference

Joohwan Ko, Justin Domke · 2026

Read Amortized Factor Inference Networks for Posterior Inference by Joohwan Ko, Justin Domke (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-26419.txt.

Description

Joohwan Ko and Justin Domke ask whether one inference network can serve many Bayesian models without retraining per family. Their Amortized Factor Inference Networks portray a model as typed prior and likelihood factors, then map that portrayal plus observations to a Gaussian posterior estimate via an encode-merge-decode network. Learned weights stay independent of latent size and observation count. Synthetic and real benchmarks show one trained network rivaling per-task variational inference and NUTS sampling at far lower test-time cost. Optional importance sampling sharpens accuracy further.

Catalog record

CREATORJoohwan Ko, Justin Domke
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE97124
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.26419

Browse the full library · About the Tower of Babel