Goto

Collaborating Authors

 bartlettetal

Uniform Convergenceof Interpolators: Gaussian Width, Norm Boundsand Benign Overfitting

Neural Information Processing Systems

"Overfitting Can Be Harmlessfor Basis Pursuit: Only toa Degree.

A High-Dimensional

Neural Information Processing Systems

Note statement, the, andC onlyneednandD. 4

Early-stoppedneuralnetworksareconsistent

Neural Information Processing Systems

Rounding out the story and contributions, firstly we present a brief toy univariate model hinting towards the necessity of early stopping: concretely, any univariate predictor satisfying alocal interpolation propertycan not achieve optimal test error for noisy distributions.