Robust regression with covariate filtering: Heavy tails and adversarial contamination

Abstract

We study the problem of linear regression where both covariates and responses are potentially (i) heavy-tailed and (ii) adversarially contaminated. Several computationally efficient estimators have...