Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Hu, Qingyi, Hodgkinson, Liam
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2602.12680
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918335790186496
author Hu, Qingyi
Hodgkinson, Liam
author_facet Hu, Qingyi
Hodgkinson, Liam
contents The rule of thumb regarding the relationship between the bias-variance tradeoff and model size plays a key role in classical machine learning, but is now well-known to break down in the overparameterized setting as per the double descent curve. In particular, minimum-norm interpolating estimators can perform well, suggesting the need for new tradeoff in these settings. Accordingly, we propose a regularization-sharpness tradeoff for overparameterized linear regression with an $\ell^p$ penalty. Inspired by the interpolating information criterion, our framework decomposes the selection penalty into a regularization term (quantifying the alignment of the regularizer and the interpolator) and a geometric sharpness term on the interpolating manifold (quantifying the effect of local perturbations), yielding a tradeoff analogous to bias-variance. Building on prior analyses that established this information criterion for ridge regularizers, this work first provides a general expression of the interpolating information criterion for $\ell^p$ regularizers where $p \ge 2$. Subsequently, we extend this to the LASSO interpolator with $\ell^1$ regularizer, which induces stronger sparsity. Empirical results on real-world datasets with random Fourier features and polynomials validate our theory, demonstrating how the tradeoff terms can distinguish performant linear interpolators from weaker ones.
format Preprint
id arxiv_https___arxiv_org_abs_2602_12680
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle A Regularization-Sharpness Tradeoff for Linear Interpolators
Hu, Qingyi
Hodgkinson, Liam
Machine Learning
The rule of thumb regarding the relationship between the bias-variance tradeoff and model size plays a key role in classical machine learning, but is now well-known to break down in the overparameterized setting as per the double descent curve. In particular, minimum-norm interpolating estimators can perform well, suggesting the need for new tradeoff in these settings. Accordingly, we propose a regularization-sharpness tradeoff for overparameterized linear regression with an $\ell^p$ penalty. Inspired by the interpolating information criterion, our framework decomposes the selection penalty into a regularization term (quantifying the alignment of the regularizer and the interpolator) and a geometric sharpness term on the interpolating manifold (quantifying the effect of local perturbations), yielding a tradeoff analogous to bias-variance. Building on prior analyses that established this information criterion for ridge regularizers, this work first provides a general expression of the interpolating information criterion for $\ell^p$ regularizers where $p \ge 2$. Subsequently, we extend this to the LASSO interpolator with $\ell^1$ regularizer, which induces stronger sparsity. Empirical results on real-world datasets with random Fourier features and polynomials validate our theory, demonstrating how the tradeoff terms can distinguish performant linear interpolators from weaker ones.
title A Regularization-Sharpness Tradeoff for Linear Interpolators
topic Machine Learning
url https://arxiv.org/abs/2602.12680