CLT in high-dimensional Bayesian linear regression with low SNR

Seunghyun Lee, Nabarun Deb, Sumit Mukherjee · 2025

Read CLT in high-dimensional Bayesian linear regression with low SNR by Seunghyun Lee, Nabarun Deb, Sumit Mukherjee (2025) 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-2507-23285.txt.

Description

Bayesian forecasts in many dimensions often refuse to shrink toward certainty, and their spread still follows a pattern. Seunghyun Lee, Nabarun Deb, and Sumit Mukherjee prove that pattern in this 2025 paper. The setting is Bayesian linear regression where signal and noise stay in balance. Neither the data nor the prior beliefs dominate, unlike the tidy cases theory usually covers.

For one-dimensional views of the posterior, the spread settles into a Gaussian shape. The center, though, sits at a simpler approximation of the posterior, not the true value. Two concrete models show the effect, one of them deliberately misspecified. Honest uncertainty bands can then be built and their coverage computed. Modern data sets live in this balanced regime.

Catalog record

CREATORSeunghyun Lee, Nabarun Deb, Sumit Mukherjee
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE216 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-23285

Browse the full library · About the Tower of Babel