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

Recent research finds CNN models for image classification demonstrateoverlapped adversarial vulnerabilities: adversarial attacks can mislead CNNmodels with small perturbations, which can effectively transfer betweendifferent models trained on the same dataset. Adversarial training, as ageneral robustness improvement technique, eliminates the vulnerability in asingle model by forcing it to learn robust features. The process is hard, oftenrequires models with large capacity, and suffers from significant loss on cleandata accuracy. Alternatively, ensemble methods are proposed to inducesub-models with diverse outputs against a transfer adversarial example, makingthe ensemble robust against transfer attacks even if each sub-model isindividually non-robust. Only small clean accuracy drop is observed in theprocess. However, previous ensemble training methods are not efficacious ininducing such diversity and thus ineffective on reaching robust ensemble. Wepropose DVERGE, which isolates the adversarial vulnerability in each sub-modelby distilling non-robust features, and diversifies the adversarialvulnerability to induce diverse outputs against a transfer attack. The noveldiversity metric and training procedure enables DVERGE to achieve higherrobustness against transfer attacks comparing to previous ensemble methods, andenables the improved robustness when more sub-models are added to the ensemble.The code of this work is available at https://github.com/zjysteven/DVERGE

DVERGE: Diversifying Vulnerabilities for Enhanced Robust Generation of Ensembles