Kernel embeddings and the separation of measure phenomenon

Leonardo V. Santoro, Kartik G. Waghmare, Victor M. Panaretos · 2026

Read Kernel embeddings and the separation of measure phenomenon by Leonardo V. Santoro, Kartik G. Waghmare, Victor M. Panaretos (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/pmc-13250597.txt.

Description

Santoro, Waghmare, and Panaretos prove a striking result about telling probability distributions apart with kernel methods. Their 2026 Proceedings of the National Academy of Sciences paper works with kernel covariance embeddings, which map a whole distribution into a point in a large mathematical space. The setting covers continuous distributions on very general spaces.

The authors show that two different distributions always map to Gaussian measures that sit on essentially separate supports, making them perfectly distinguishable in principle. A classical result on Gaussian measures, the Feldman-Hájek dichotomy, drives the proof. Even a tiny change to a distribution becomes maximally visible after this embedding. They call this the separation of measure phenomenon and suggest it explains why kernel methods work so well in practice.

Catalog record

CREATORLeonardo V. Santoro, Kartik G. Waghmare, Victor M. Panaretos
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC13250597

Browse the full library · About the Tower of Babel