Matches in SemOpenAlex for { <https://semopenalex.org/work/W3168907554> ?p ?o ?g. }
Showing items 1 to 80 of
80
with 100 items per page.
- W3168907554 abstract "Despite their overwhelming success on a wide range of applications, convolutional neural networks (CNNs) are widely recognized to be vulnerable to adversarial examples. This intriguing phenomenon led to a competition between adversarial attacks and defense techniques. So far, adversarial training is the most widely used method for defending against adversarial attacks. It has also been extended to defend against universal adversarial perturbations (UAPs). The SOTA universal adversarial training (UAT) method optimizes a single perturbation for all training samples in the mini-batch. In this work, we find that a UAP does not attack all classes equally. Inspired by this observation, we identify it as the source of the model having unbalanced robustness. To this end, we improve the SOTA UAT by proposing to utilize class-wise UAPs during adversarial training. On multiple benchmark datasets, our class-wise UAT leads superior performance for both clean accuracy and adversarial robustness against universal attack." @default.
- W3168907554 created "2021-06-22" @default.
- W3168907554 creator A5000590564 @default.
- W3168907554 creator A5012455275 @default.
- W3168907554 creator A5024432805 @default.
- W3168907554 creator A5057230698 @default.
- W3168907554 date "2021-07-05" @default.
- W3168907554 modified "2023-10-16" @default.
- W3168907554 title "Universal Adversarial Training with Class-Wise Perturbations" @default.
- W3168907554 cites W2194775991 @default.
- W3168907554 cites W2243397390 @default.
- W3168907554 cites W2543927648 @default.
- W3168907554 cites W2885039223 @default.
- W3168907554 cites W2962700793 @default.
- W3168907554 cites W2963165363 @default.
- W3168907554 cites W2963855547 @default.
- W3168907554 cites W2964137095 @default.
- W3168907554 cites W2971848556 @default.
- W3168907554 cites W2997223206 @default.
- W3168907554 cites W2997502936 @default.
- W3168907554 cites W2998421476 @default.
- W3168907554 cites W3009394976 @default.
- W3168907554 cites W3034643863 @default.
- W3168907554 cites W3119029226 @default.
- W3168907554 cites W3120903390 @default.
- W3168907554 cites W3190041646 @default.
- W3168907554 doi "https://doi.org/10.1109/icme51207.2021.9428419" @default.
- W3168907554 hasPublicationYear "2021" @default.
- W3168907554 type Work @default.
- W3168907554 sameAs 3168907554 @default.
- W3168907554 citedByCount "8" @default.
- W3168907554 countsByYear W31689075542021 @default.
- W3168907554 countsByYear W31689075542022 @default.
- W3168907554 countsByYear W31689075542023 @default.
- W3168907554 crossrefType "proceedings-article" @default.
- W3168907554 hasAuthorship W3168907554A5000590564 @default.
- W3168907554 hasAuthorship W3168907554A5012455275 @default.
- W3168907554 hasAuthorship W3168907554A5024432805 @default.
- W3168907554 hasAuthorship W3168907554A5057230698 @default.
- W3168907554 hasBestOaLocation W31689075542 @default.
- W3168907554 hasConcept C104317684 @default.
- W3168907554 hasConcept C108583219 @default.
- W3168907554 hasConcept C119857082 @default.
- W3168907554 hasConcept C154945302 @default.
- W3168907554 hasConcept C185592680 @default.
- W3168907554 hasConcept C2777212361 @default.
- W3168907554 hasConcept C2984842247 @default.
- W3168907554 hasConcept C37736160 @default.
- W3168907554 hasConcept C41008148 @default.
- W3168907554 hasConcept C55493867 @default.
- W3168907554 hasConcept C63479239 @default.
- W3168907554 hasConceptScore W3168907554C104317684 @default.
- W3168907554 hasConceptScore W3168907554C108583219 @default.
- W3168907554 hasConceptScore W3168907554C119857082 @default.
- W3168907554 hasConceptScore W3168907554C154945302 @default.
- W3168907554 hasConceptScore W3168907554C185592680 @default.
- W3168907554 hasConceptScore W3168907554C2777212361 @default.
- W3168907554 hasConceptScore W3168907554C2984842247 @default.
- W3168907554 hasConceptScore W3168907554C37736160 @default.
- W3168907554 hasConceptScore W3168907554C41008148 @default.
- W3168907554 hasConceptScore W3168907554C55493867 @default.
- W3168907554 hasConceptScore W3168907554C63479239 @default.
- W3168907554 hasLocation W31689075541 @default.
- W3168907554 hasLocation W31689075542 @default.
- W3168907554 hasOpenAccess W3168907554 @default.
- W3168907554 hasPrimaryLocation W31689075541 @default.
- W3168907554 hasRelatedWork W2950183588 @default.
- W3168907554 hasRelatedWork W2952919291 @default.
- W3168907554 hasRelatedWork W3094843325 @default.
- W3168907554 hasRelatedWork W3193857078 @default.
- W3168907554 hasRelatedWork W3207178610 @default.
- W3168907554 hasRelatedWork W3208304128 @default.
- W3168907554 hasRelatedWork W4306177400 @default.
- W3168907554 hasRelatedWork W4311734044 @default.
- W3168907554 hasRelatedWork W4322759769 @default.
- W3168907554 hasRelatedWork W4365420856 @default.
- W3168907554 isParatext "false" @default.
- W3168907554 isRetracted "false" @default.
- W3168907554 magId "3168907554" @default.
- W3168907554 workType "article" @default.