no code implementations • 1 Dec 2021 • Shunan Yao, Bradley Rava, Xin Tong, Gareth James
It is somewhat surprising that even when common NP classifiers ignore the label noise in the training stage, they are still able to control the type I error with high probability.