On the Generalization of Models Trained with SGD: Information-Theoretic Bounds and Implications.

Ziqiao Wang, Yongyi Mao

Browse the full ICLR paper archive.