Overview

Robust regression studies how to estimate relationships between variables when data contain outliers, contamination, heavy-tailed noise, or distribution shifts. Common approaches include robust losses such as Huber and Welsch, trimming and median-of-means estimators, sparse thresholding, and distributionally robust optimization. The field also develops guarantees for settings such as adversarial corruption, missing features, and multi-task learning, alongside practical methods for uncertainty estimates and nonlinear models. A central tension is balancing protection against bad data with statistical efficiency and computational cost, since robust methods do not always outperform ordinary least squares when its assumptions are adequate.

Latest papers

CardsList

34 papers3 papers in the last four weeks, down 25% on the four weeks before. 0.0% of all new papers.

  1. Outlier-Robust Multi-Output Gaussian Processes

    Oct 30, 2025Joshua Rooijakkers, Leiv Rønneberg, François-Xavier Briol +2Gaussian ProcessesMulti-Output Gaussian Processes

  2. Linear Regression under Missing or Corrupted Coordinates

    Sep 23, 2025Ilias Diakonikolas, Jelena Diakonikolas, Daniel M. Kane +2Robust RegressionLearning with Missing Data

  3. Robust Linear Predictions: Analyses of Uniform Concentration, Fast Rates and Model Misspecification

    Jan 6, 2022Saptarshi Chakraborty, Debolina Paul, Swagatam DasRobust RegressionModel Misspecification