Cross Validation for Correlated Data in Regression and Classification Models, with Applications to Deep Learning

Oren Yuval · 2025

Read Cross Validation for Correlated Data in Regression and Classification Models, with Applications to Deep Learning by Oren Yuval (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-2502-14808.txt.

Description

Oren Yuval and Saharon Rosset present, in a 2025 paper, a fix for cross-validation when data points lean on each other. Standard cross-validation assumes each record stands alone. Clustered, spatial, and time-linked data break that rule, so test scores can mislead model choice.

The authors write the bias between the standard score and the true error as a single term. They then give data-based ways to estimate that term for any model, even deep networks. Adding the estimate back yields a corrected score. The idea grew from a linear model case and now covers wide regression and classification tasks. Tests on fake and real data with tangled structure show the corrected score tracks true error better than the standard one.

Catalog record

CREATOROren Yuval
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE79 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-14808

Browse the full library · About the Tower of Babel