Publication

Robustness in deep learning: The good (width), the bad (depth), and the ugly (initialization)