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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

In this work, we investigate the phenomenon that robust image classifiershave human-recognizable features -- often referred to as interpretability -- asrevealed through the input gradients of their score functions and theirsubsequent adversarial perturbations. In particular, we demonstrate thatstate-of-the-art methods for adversarial training incorporate two terms -- onethat orients the decision boundary via minimizing the expected loss, andanother that induces smoothness of the classifier's decision surface bypenalizing the local Lipschitz constant. Through this demonstration, we providea unified discussion of gradient and Jacobian-based regularizers that have beenused to encourage adversarial robustness in prior works. Following thisdiscussion, we give qualitative evidence that the coupling of smoothness andorientation of the decision boundary is sufficient to induce the aforementionedhuman-recognizability phenomenon.

On the human-recognizability phenomenon of adversarially trained deep image classifiers