🤖 AI Summary
This paper investigates statistical estimation under Wasserstein-$r$ contamination ($r geq 1$) with $ell_q$-norm constraints ($q in [1,infty]$), addressing mean estimation, linear regression, and nonparametric density estimation. Unlike Huber-type sparse contamination models, this setting permits adversarial perturbations of all observations at bounded transportation cost, distinguishing independent and joint (coordinated) contamination regimes. Leveraging Wasserstein-type distributional shift modeling, the analysis integrates Le Cam’s, Fano’s, and Assouad’s methods, modulus-of-continuity arguments, and prior sequence constructions to establish, for the first time, a unified minimax risk framework under $ell_q^r$ loss. Key contributions include: (i) exact minimax risk characterizations and optimality proofs for mean and linear regression, confirming the minimax optimality of the sample mean and least squares estimators; (ii) a novel analytical toolkit tailored to generalized distributional shifts; and (iii) tight finite-sample bounds alongside nearly optimal estimators.
📝 Abstract
Distribution shifts are a serious concern in modern statistical learning as they can systematically change the properties of the data away from the truth. We focus on Wasserstein distribution shifts, where every data point may undergo a slight perturbation, as opposed to the Huber contamination model where a fraction of observations are outliers. We consider perturbations that are either independent or coordinated joint shifts across data points. We analyze several important statistical problems, including location estimation, linear regression, and non-parametric density estimation. Under a squared loss for mean estimation and prediction error in linear regression, we find the exact minimax risk, a least favorable perturbation, and show that the sample mean and least squares estimators are respectively optimal. For other problems, we provide nearly optimal estimators and precise finite-sample bounds. We also introduce several tools for bounding the minimax risk under general distribution shifts, not just for Wasserstein perturbations, such as a smoothing technique for location families, and generalizations of classical tools including least favorable sequences of priors, the modulus of continuity, as well as Le Cam's, Fano's, and Assouad's methods.
