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

With the great success of deep neural networks, adversarial learning hasreceived widespread attention in various studies, ranging from multi-classlearning to multi-label learning. However, existing adversarial attacks towardmulti-label learning only pursue the traditional visual imperceptibility butignore the new perceptible problem coming from measures such as Precision@$k$and mAP@$k$. Specifically, when a well-trained multi-label classifier performsfar below the expectation on some samples, the victim can easily realize thatthis performance degeneration stems from attack, rather than the model itself.Therefore, an ideal multi-labeling adversarial attack should manage to not onlydeceive visual perception but also evade monitoring of measures. To this end,this paper first proposes the concept of measure imperceptibility. Then, anovel loss function is devised to generate such adversarial perturbations thatcould achieve both visual and measure imperceptibility. Furthermore, anefficient algorithm, which enjoys a convex objective, is established tooptimize this objective. Finally, extensive experiments on large-scalebenchmark datasets, such as PASCAL VOC 2012, MS COCO, and NUS WIDE, demonstratethe superiority of our proposed method in attacking the top-$k$ multi-labelsystems.

When Measures are Unreliable: Imperceptible Adversarial Perturbations toward Top-$k$ Multi-Label Learning