Solve for the Hyperparameter, Skip the Search: Kolmogorov-Optimal Scaling Laws for Spline Regression

Yong Yi Bay Kathleen A. Yearick · 2026

Read Solve for the Hyperparameter, Skip the Search: Kolmogorov-Optimal Scaling Laws for Spline Regression by Yong Yi Bay Kathleen A. Yearick (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-2606-23575.txt.

Introduction

Bay and Yearick show that the key tuning knob of spline regression can be computed directly instead of searched, in a 2026 paper on statistical learning. Spline regression fits smooth curves built from pieces, and its resolution setting controls how wiggly the curve may be. Standard practice tries every grid value and keeps the winner by cross validation, wasting computation.

Classical approximation theory pins down how bias shrinks as resolution grows, at exactly the best rate any linear method can reach for the smoothness class. A shortcut identity also gives the leave one out error of a fit from that single fit alone. Balancing the two known curves yields the best resolution in closed form. Their algorithm, KORE, needs only about a dozen fits, yet matches full grid searches while fitting roughly eight times fewer models.

Catalog record

CREATORYong Yi Bay Kathleen A. Yearick
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE142 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.23575

Browse the full library · About the Tower of Babel