Abstract:As one of the triumphs and milestones of robust statistics, Huber regression plays an important role in robust inference and estimation. It has also been finding a great variety of applications in machine learning. In a parametric setup, it has been extensively studied. However, in the statistical learning context where a function is typically learned in a nonparametric way, there is still a lack of theoretical understanding of how Huber regression estimators learn the conditional mean function and why it works in the absence of light-tailed noise assumptions. To address these fundamental questions, we conduct an assessment of Huber regression from a statistical learning viewpoint. First, we show that the usual risk consistency property of Huber regression estimators, which is usually pursued in machine learning, cannot guarantee their learnability in mean regression. Second, we argue that Huber regression should be implemented in an adaptive way to perform mean regression, implying that one needs to tune the scale parameter in accordance with the sample size and the moment condition of the noise. Third, with an adaptive choice of the scale parameter, we demonstrate that Huber regression estimators can be asymptotic mean regression calibrated under $(1+\epsilon)$-moment conditions ($\epsilon>0$). Last but not least, under the same moment conditions, we establish almost sure convergence rates for Huber regression estimators. Note that the $(1+\epsilon)$-moment conditions accommodate the special case where the response variable possesses infinite variance and so the established convergence rates justify the robustness feature of Huber regression estimators. In the above senses, the present study provides a systematic statistical learning assessment of Huber regression estimators and justifies their merits in terms of robustness from a theoretical viewpoint.

Mean estimation and regression under heavy-tailed distributions--a survey

Robust Estimation and Shrinkage in Ultrahigh Dimensional Expectile Regression with Heavy Tails and Variance Heterogeneity

Semiparametric Expectile Regression for High-dimensional Heavy-tailed and Heterogeneous Data

On mean and/or variance mixtures of normal distributions

Mean Estimation in Banach Spaces Under Infinite Variance and Martingale Dependence

On deviation probabilities in non-parametric regression with heavy-tailed noise

\Ell_1-Regression with Heavy-tailed Distributions.

A NEW PERSPECTIVE ON ROBUST M-ESTIMATION: FINITE SAMPLE THEORY AND APPLICATIONS TO DEPENDENCE-ADJUSTED MULTIPLE TESTING

Estimating the mean of a heavy-tailed distribution under random censoring

L G ] 2 5 O ct 2 01 8 l 1-regression with Heavy-tailed Distributions

High-dimensional robust regression under heavy-tailed data: Asymptotics and Universality

Heavy-Tailed Distributions and Their Applications

Concentration Inequalities for General Functions of Heavy-Tailed Random Variables

Sub-Weibull distributions: generalizing sub-Gaussian and sub-Exponential properties to heavier-tailed distributions

On Tail Decay Rate Estimation of Loss Function Distributions

Mean-Square Analysis of Discretized Itô Diffusions for Heavy-tailed Sampling

Estimating Extreme Value Index by Subsampling for Massive Datasets with Heavy-Tailed Distributions

Catoni-style confidence sequences for heavy-tailed mean estimation

Heavy-Tailed Processes for Selective Shrinkage

Beyond Catoni: Sharper Rates for Heavy-Tailed and Robust Mean Estimation

A Statistical Learning Assessment of Huber Regression