Learning Curves and Benign Overfitting of Spectral Algorithms in Large Dimensions

Weihao Lu, Qian Lin, Yingcun Xia, Dongming Huang · 2026

Read Learning Curves and Benign Overfitting of Spectral Algorithms in Large Dimensions by Weihao Lu, Qian Lin, Yingcun Xia, Dongming Huang (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-2604-23212.txt.

Introduction

Lu and three colleagues characterise learning curves and benign overfitting of spectral algorithms with comparable sample size and dimension. For kernels on the sphere, they describe excess risk across the regularisation path under source conditions for regression smoothness. The curve is not U-shaped but splits into over-regularised, under-regularised, and interpolation regimes. Benign overfitting appears across under-regularised and interpolation regimes when smoothness is positive but below a critical threshold. Kernel curves in the regularised regime follow a sequence model.

Catalog record

CREATORWeihao Lu, Qian Lin, Yingcun Xia, Dongming Huang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE262522
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.23212

Browse the full library · About the Tower of Babel