On the robustness of minimum norm interpolators and regularized empirical risk minimizers
This work provides a theoretical understanding of the robustness of overparameterized models for researchers working on machine learning theory, especially in the presence of adversarial noise.
This paper develops a theory for minimum norm interpolating estimators and regularized empirical risk minimizers in linear models with additive errors, even adversarial ones. It provides a quantitative bound for prediction error and shows that for sparsity or low-rank inducing norms, the prediction error is on the order of the average noise level, given sufficient overparameterization and small regularization.
This article develops a general theory for minimum norm interpolating estimators and regularized empirical risk minimizers (RERM) in linear models in the presence of additive, potentially adversarial, errors. In particular, no conditions on the errors are imposed. A quantitative bound for the prediction error is given, relating it to the Rademacher complexity of the covariates, the norm of the minimum norm interpolator of the errors and the size of the subdifferential around the true parameter. The general theory is illustrated for Gaussian features and several norms: The $\ell_1$, $\ell_2$, group Lasso and nuclear norms. In case of sparsity or low-rank inducing norms, minimum norm interpolators and RERM yield a prediction error of the order of the average noise level, provided that the overparameterization is at least a logarithmic factor larger than the number of samples and that, in case of RERM, the regularization parameter is small enough. Lower bounds that show near optimality of the results complement the analysis.