On the Inconsistency of Kernel Ridgeless Regression in Fixed Dimensions

DOI10.1137/22M1499819MaRDI QIDQ6070298zbMATH OpenFDO

Authors Mikhail Belkin, Parthe Pandit

Publication date 20 November 2023

Published in SIAM Journal on Mathematics of Data Science (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2205.13525

zbMATH Keywords

benign overfittingconsistencyinterpolationkernel machinesnonparametric regressionridgeless regression

Mathematics Subject Classification ID

Nonparametric estimation (62G05)

Abstract: ``Benign overfitting, the ability of certain algorithms to interpolate noisy training data and yet perform well out-of-sample, has been a topic of considerable recent interest. We show, using a fixed design setup, that an important class of predictors, kernel machines with translation-invariant kernels, does not exhibit benign overfitting in fixed dimensions. In particular, the estimated predictor does not converge to the ground truth with increasing sample size, for any non-zero regression function and any (even adaptive) bandwidth selection. To prove these results, we give exact expressions for the generalization error, and its decomposition in terms of an approximation error and an estimation error that elicits a trade-off based on the selection of the kernel bandwidth. Our results apply to commonly used translation-invariant kernels such as Gaussian, Laplace, and Cauchy.

Cites work

• A distribution-free theory of nonparametric regression
• Analysis.
• Benign overfitting in linear regression
• Generalization error of random feature and kernel methods: hypercontractivity and kernel matrix concentration
• scientific article; zbMATH DE number 1804115 (Why is no real title available?)
• scientific article; zbMATH DE number 1375577 (Why is no real title available?)
• Just interpolate: kernel ``ridgeless regression can generalize
• Reconciling modern machine-learning practice and the classical bias-variance trade-off
• Surprises in high-dimensional ridgeless least squares interpolation
• The Generalization Error of Random Features Regression: Precise Asymptotics and the Double Descent Curve
• The Hilbert kernel regression estimate.
• Two models of double descent for weak features

Cited in

(4)

• Rows versus Columns: Randomized Kaczmarz or Gauss--Seidel for Ridge Regression
• The phase diagram of kernel interpolation in large dimensions
• Optimal rates of kernel ridge regression under source condition in large dimensions
• Sobolev norm inconsistency of kernel interpolation

This page was built for publication: On the Inconsistency of Kernel Ridgeless Regression in Fixed Dimensions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6070298)