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 response to the threat of adversarial examples, adversarial trainingprovides an attractive option for enhancing the model robustness by trainingmodels on online-augmented adversarial examples. However, most of the existingadversarial training methods focus on improving the robust accuracy bystrengthening the adversarial examples but neglecting the increasing shiftbetween natural data and adversarial examples, leading to a dramatic decreasein natural accuracy. To maintain the trade-off between natural and robustaccuracy, we alleviate the shift from the perspective of feature adaption andpropose a Feature Adaptive Adversarial Training (FAAT) optimizing theclass-conditional feature adaption across natural data and adversarialexamples. Specifically, we propose to incorporate a class-conditionaldiscriminator to encourage the features become (1) class-discriminative and (2)invariant to the change of adversarial attacks. The novel FAAT frameworkenables the trade-off between natural and robust accuracy by generatingfeatures with similar distribution across natural and adversarial data, andachieve higher overall robustness benefited from the class-discriminativefeature characteristics. Experiments on various datasets demonstrate that FAATproduces more discriminative features and performs favorably againststate-of-the-art methods. Codes are available athttps://github.com/VisionFlow/FAAT.

Push Stricter to Decide Better: A Class-Conditional Feature Adaptive Framework for Improving Adversarial Robustness