Noisy Early Stopping for Noisy Labels

William Toner, Amos Storkey

Training neural network classifiers on datasets contaminated with noisy labels significantly increases the risk of overfitting. Thus, effectively implementing Early Stopping in noi…