Empirical risk minimization for heavy-tailed losses

DOI10.1214/15-AOS1350zbMath1326.62066arXiv1406.2462OpenAlexW3104898220WikidataQ105584309 ScholiaQ105584309MaRDI QIDQ892246

Gábor Lugosi, Emilien Joly, Christian Brownlees

Publication date: 18 November 2015

Published in: The Annals of Statistics (Search for Journal in Brave)

Full work available at URL: https://arxiv.org/abs/1406.2462

zbMATH Keywords

robust regressionempirical risk minimizationheavy-tailed dataCatoni's estimatorrobust \(k\)-means clustering

Mathematics Subject Classification ID

Asymptotic properties of parametric estimators (62F12) Asymptotic properties of nonparametric inference (62G20) Classification and discrimination; cluster analysis (statistical aspects) (62H30) Robustness and adaptive procedures (parametric inference) (62F35) Statistics of extreme values; tail inference (62G32)

Related Items

On least squares estimation under heteroscedastic and heavy-tailed errors, A shrinkage principle for heavy-tailed data: high-dimensional robust low-rank matrix recovery, Robust Bregman clustering, A New Principle for Tuning-Free Huber Regression, Distributionally-robust machine learning using locally differentially-private data, Fast rates of minimum error entropy with heavy-tailed noise, Posterior concentration and fast convergence rates for generalized Bayesian learning, Regularization, sparse recovery, and median-of-means tournaments, Learning with risks based on M-location, Robust inference via multiplier bootstrap, Robust supervised learning with coordinate gradient descent, Robust high-dimensional tuning free multiple testing, Robust regression using biased objectives, Understanding Implicit Regularization in Over-Parameterized Single Index Model, Robust classification via MOM minimization, Adaptive Huber Regression, Sub-Gaussian estimators of the mean of a random matrix with heavy-tailed entries, Learning from MOM's principles: Le Cam's approach, Unnamed Item, Data-driven distributionally robust optimization using the Wasserstein metric: performance guarantees and tractable reformulations, User-friendly covariance estimation for heavy-tailed distributions, Robust \(k\)-means clustering for distributions with two moments, Convergence rates of least squares regression estimators with heavy-tailed errors, Efficient learning with robust gradient descent, Learning under \((1 + \epsilon)\)-moment conditions, Mean estimation and regression under heavy-tailed distributions: A survey, High dimensional generalized linear models for temporal dependent data, Distribution-free robust linear regression

Cites Work

• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Unnamed Item
• Geometric median and robust estimation in Banach spaces
• Fast rates for empirical vector quantization
• Robust linear least squares regression
• Concentration inequalities and model selection. Ecole d'Eté de Probabilités de Saint-Flour XXXIII -- 2003.
• Strong consistency of k-means clustering
• A central limit theorem for k-means clustering
• Central limit theorems for empirical measures
• The space complexity of approximating the frequency moments
• Weak convergence and empirical processes. With applications to statistics
• Challenging the empirical mean and empirical variance: a deviation study
• Local Rademacher complexities and oracle inequalities in risk minimization. (2004 IMS Medallion Lecture). (With discussions and rejoinder)
• Empirical minimization
• Learning without Concentration
• Convergence of Vector Quantizers with Applications to Optimal Quantization
• Theory of Classification: a Survey of Some Recent Advances
• Improved Minimax Bounds on the Test and Training Distortion of Empirically Designed Vector Quantizers
• Individual Convergence Rates in Empirical Vector Quantizer Design
• On the Performance of Clustering in Hilbert Spaces
• Quantization and the method of<tex>k</tex>-means
• The Generic Chaining
• The minimax distortion redundancy in empirical quantizer design
• $K$-Dimensional Coding Schemes in Hilbert Spaces
• Bandits With Heavy Tail