Finding Biological Plausibility for Adversarially Robust Features via Metameric Tasks

Anne Harrington, Arturo Deza

Abstract

Recent work suggests that feature constraints in the training datasets of deep neural networks (DNNs) drive robustness to adversarial noise (Ilyas et al., 2019). The representations learned by such...
BibTeX
@inproceedings{harrington2022finding,
  author = {Anne Harrington and Arturo Deza},
  title = {{Finding Biological Plausibility for Adversarially Robust Features via Metameric Tasks}},
  booktitle = {International Conference on Learning Representations},
  year = {2022},
  url = {https://openreview.net/forum?id=-vaqbFtifz},
}
Figure