3 ms·
Adversarially robust classifiers have interpretable gradients and feature representations [0]. The problem seems to be that the standard networks capture all th
by red75prime 8y ago
Adversarially robust classifiers have interpretable gradients and feature representations [0]. The problem seems to be that the standard networks capture all the statistics there is, including surface statistical regularities and noise. It can be mitigated, though.
[0]: https://arxiv.org/abs/1805.12152 https://arxiv.org/abs/1805.12152
- mlthoughts2018 8y agoThanks for the link! Note that it’s not just adversarial robustness that is a problem, as the paper’s result merely with altering surface statistics with a Fourier domain filter applied to the training data already creates problems for interpreting the network’s internal representation as having any type of semantic representation of the underlying structure relevant for the task, without needing to involve the part about adversarial robustness at all.