Selective Nonparametric Regression via Testing

Fedor Noskov, Alexander Fishkov, Maxim Panov

Prediction with the possibility of abstention (or selective prediction) is an important problem for error-critical machine learning applications. While well-studied in the classifi…