Tighter sparse variational Gaussian processes

Thang D. Bui [email protected], Matthew Ashman [email protected], Richard E. Turner [email protected] · 2025

Read Tighter sparse variational Gaussian processes by Thang D. Bui [email protected], Matthew Ashman [email protected], Richard E. Turner [email protected] (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-2502-04750.txt.

Description

Gaussian processes scale to big data through a sparse shortcut, and a 2025 arXiv paper by Thang Bui, Matthew Ashman, and Richard Turner proves the shortcut can be tightened for free. The standard version forces the approximate posterior, given a small set of summary points, to match the prior at the training inputs.

Letting those variances shrink instead yields a bound that is provably closer to the truth. The authors derive the fast collapsed form for regression and extend the trick to structured point sets and latent variable models. Across benchmarks the tighter version matched or beat the standard one at the same cost.

Catalog record

CREATORThang D. Bui [email protected], Matthew Ashman [email protected], Richard E. Turner [email protected]
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-04750

Browse the full library · About the Tower of Babel