Conformal Inference is (almost) Free for Neural Networks Trained with Early Stopping.

Ziyi Liang, Yanfei Zhou, Matteo Sesia

Browse the full ICML paper archive.