English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Adversarial training is a training scheme designed to counter adversarialattacks by augmenting the training dataset with adversarial examples.Surprisingly, several studies have observed that loss gradients fromadversarially trained DNNs are visually more interpretable than those fromstandard DNNs. Although this phenomenon is interesting, there are only fewworks that have offered an explanation. In this paper, we attempted to bridgethis gap between adversarial robustness and gradient interpretability. To thisend, we identified that loss gradients from adversarially trained DNNs alignbetter with human perception because adversarial training restricts gradientscloser to the image manifold. We then demonstrated that adversarial trainingcauses loss gradients to be quantitatively meaningful. Finally, we showed thatunder the adversarial training framework, there exists an empirical trade-offbetween test accuracy and loss gradient interpretability and proposed twopotential approaches to resolving this trade-off.

Bridging Adversarial Robustness and Gradient Interpretability