Does data interpolation contradict statistical optimality?

06/25/2018

∙

by   Mikhail Belkin, et al.

∙

4

∙

share

We show that learning methods interpolating the training data can achieve optimal rates for the problems of nonparametric regression and prediction with square loss.