Scalable Gaussian process inference via neural feature maps

Anthony Stephenson · 2026

Read Scalable Gaussian process inference via neural feature maps by Anthony Stephenson (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-2605-10285.txt.

Description

Anthony Stephenson offers a faster route to Gaussian process inference, a Bayesian method prized for flexible modeling with honest uncertainty estimates. The classic drawback is cubic scaling: naive inference grows with the cube of the training-set size, quickly becoming infeasible. His 2026 paper builds expressive kernels from neural feature maps instead.

The learned feature map can be read as an optimal low-rank approximation to a Gram matrix from an implied reproducing kernel Hilbert space, which grounds the posterior's consistency. Spectral analysis of the induced kernels leads to product feature-map kernels that counter oversmoothing. The result is fast, scalable, exact inference with little upfront work, applied to regression and classification across tabular and structured data. On benchmarks it beat existing methods in accuracy and in training and prediction efficiency.

Catalog record

CREATORAnthony Stephenson
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE124 KB
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2605.10285

Browse the full library · About the Tower of Babel