A_Closer_Look_at_Double_Backpropagation.md

@article{2020arXiv200301895S,
archivePrefix = {arXiv},
arxivId = {cs.CV/2003.01895},
author = {Sun, Chengjin and Chen, Sizhe and Huang, Xiaolin},
eprint = {2003.01895},
journal = {arXiv e-prints},
keywords = {Computer Science - Computer Vision and Pattern Rec,Computer Science - Machine Learning},
month = {mar},
pages = {arXiv:2003.01895},
primaryClass = {cs.CV},
title = {{Double Backpropagation for Training Autoencoders against Adversarial Attack}},
year = {2020}
}

Double Backpropagation

$$\ell(f(x), y)+\lambda \cdot\left|\nabla_{x} \ell(f(x), y)\right|_{2}^{2}$$

Frobenius Norm, similar to l2 norm for vectors.

Too much mathmatics for me to summarize.