On adaptive estimation using the sup-norm losses

Abstract

We consider the problem of recovering smooth functions from noisy data using the supnorm as the quality criterion Starting with a natural projection estimator we show a datadriven procedure to be adaptive asymptotically minimax