A new measure for overfitting and its implications for backdooring of deep learning

Open in new window