Test Set Sizing for Regularized Models

Alexander Dubbs · 2025

Read Test Set Sizing for Regularized Models by Alexander Dubbs (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-2504-19231.txt.

Description

Alexander Dubbs works out the best split of data into training and test parts for ridge regression, in a paper from two thousand twenty-five. Ridge regression shrinks weights to curb overfit, and its tuning strength might sway the best split. The math is done in the large-data limit with the feature count held fixed.

The tuning strength turns out to matter only weakly and can be dropped in the limit. The best split then rests on row count and feature count alone. It matches the known plain regression split to the first two terms. The aim is honest error reads, as close as can be to true skill. Readers get a first exact split rule of its kind.

Catalog record

CREATORAlexander Dubbs
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE20 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-19231

Browse the full library · About the Tower of Babel