RobustAugMix: Joint Optimization of Natural and Adversarial Robustness

Abstract

Machine learning models often suffer performance degradation when faced with corrupted data. In this work, we explore a technique that combines a data augmentation strategy (AugMix) with adversarial training, in order to increase robustness to both natural and adversarial forms of data corruption.

PDF NeurIPSW OpenReview Semantic Scholar

Cite

Text

Martinez-Martinez and Brown. "RobustAugMix: Joint Optimization of Natural and Adversarial Robustness." NeurIPS 2022 Workshops: MLSW, 2022.

Markdown

[Martinez-Martinez and Brown. "RobustAugMix: Joint Optimization of Natural and Adversarial Robustness." NeurIPS 2022 Workshops: MLSW, 2022.](https://mlanthology.org/neuripsw/2022/martinezmartinez2022neuripsw-robustaugmix/)

BibTeX

@inproceedings{martinezmartinez2022neuripsw-robustaugmix,
  title     = {{RobustAugMix: Joint Optimization of Natural and Adversarial Robustness}},
  author    = {Martinez-Martinez, Josue and Brown, Olivia},
  booktitle = {NeurIPS 2022 Workshops: MLSW},
  year      = {2022},
  url       = {https://mlanthology.org/neuripsw/2022/martinezmartinez2022neuripsw-robustaugmix/}
}