FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2996344901> ?p ?o ?g. }

• W2996344901 abstract "Deep neural networks (DNNs) are vulnerable to adversarial examples crafted by imperceptible perturbations. A range of defense techniques have been proposed to improve DNN robustness to adversarial examples, among which adversarial training has been demonstrated to be the most effective. Adversarial training is often formulated as a min-max optimization problem, with the inner maximization for generating adversarial examples. However, there exists a simple, yet easily overlooked fact that adversarial examples are only defined on correctly classified (natural) examples, but inevitably, some (natural) examples will be misclassified during training. In this paper, we investigate the distinctive influence of misclassified and correctly classified examples on the final robustness of adversarial training. Specifically, we find that misclassified examples indeed have a significant impact on the final robustness. More surprisingly, we find that different maximization techniques on misclassified examples may have a negligible influence on the final robustness, while different minimization techniques are crucial. Motivated by the above discovery, we propose a new defense algorithm called {em Misclassification Aware adveRsarial Training} (MART), which explicitly differentiates the misclassified and correctly classified examples during the training. We also propose a semi-supervised extension of MART, which can leverage the unlabeled data to further improve the robustness. Experimental results show that MART and its variant could significantly improve the state-of-the-art adversarial robustness." @default.
• W2996344901 created "2019-12-26" @default.
• W2996344901 creator A5004999983 @default.
• W2996344901 creator A5030837133 @default.
• W2996344901 creator A5031013029 @default.
• W2996344901 creator A5051448391 @default.
• W2996344901 creator A5078711649 @default.
• W2996344901 creator A5085848346 @default.
• W2996344901 date "2020-04-30" @default.
• W2996344901 modified "2023-10-01" @default.
• W2996344901 title "Improving Adversarial Robustness Requires Revisiting Misclassified Examples" @default.
• W2996344901 cites W1883420340 @default.
• W2996344901 cites W2112796928 @default.
• W2996344901 cites W2119112357 @default.
• W2996344901 cites W2145607950 @default.
• W2996344901 cites W2194775991 @default.
• W2996344901 cites W2401231614 @default.
• W2996344901 cites W2460937040 @default.
• W2996344901 cites W2590523583 @default.
• W2996344901 cites W2593910181 @default.
• W2996344901 cites W2594867206 @default.
• W2996344901 cites W2603766943 @default.
• W2996344901 cites W2618043096 @default.
• W2996344901 cites W2774018344 @default.
• W2996344901 cites W2786104118 @default.
• W2996344901 cites W2787708942 @default.
• W2996344901 cites W2787733970 @default.
• W2996344901 cites W2791953061 @default.
• W2996344901 cites W2884491628 @default.
• W2996344901 cites W2903483905 @default.
• W2996344901 cites W2907518016 @default.
• W2996344901 cites W2911634294 @default.
• W2996344901 cites W2913266441 @default.
• W2996344901 cites W2913848079 @default.
• W2996344901 cites W2945793108 @default.
• W2996344901 cites W2947294642 @default.
• W2996344901 cites W2947775933 @default.
• W2996344901 cites W2950095299 @default.
• W2996344901 cites W2951735139 @default.
• W2996344901 cites W2952477728 @default.
• W2996344901 cites W2962972504 @default.
• W2996344901 cites W2963001136 @default.
• W2996344901 cites W2963158386 @default.
• W2996344901 cites W2963207607 @default.
• W2996344901 cites W2963249138 @default.
• W2996344901 cites W2963341956 @default.
• W2996344901 cites W2963744840 @default.
• W2996344901 cites W2963857521 @default.
• W2996344901 cites W2964082701 @default.
• W2996344901 cites W2964116600 @default.
• W2996344901 cites W2964153729 @default.
• W2996344901 cites W2964197269 @default.
• W2996344901 cites W2964253222 @default.
• W2996344901 cites W2964276371 @default.
• W2996344901 cites W2970680991 @default.
• W2996344901 cites W2971316968 @default.
• W2996344901 cites W3118608800 @default.
• W2996344901 cites W3217549878 @default.
• W2996344901 hasPublicationYear "2020" @default.
• W2996344901 type Work @default.
• W2996344901 sameAs 2996344901 @default.
• W2996344901 citedByCount "118" @default.
• W2996344901 countsByYear W29963449012019 @default.
• W2996344901 countsByYear W29963449012020 @default.
• W2996344901 countsByYear W29963449012021 @default.
• W2996344901 countsByYear W29963449012022 @default.
• W2996344901 crossrefType "proceedings-article" @default.
• W2996344901 hasAuthorship W2996344901A5004999983 @default.
• W2996344901 hasAuthorship W2996344901A5030837133 @default.
• W2996344901 hasAuthorship W2996344901A5031013029 @default.
• W2996344901 hasAuthorship W2996344901A5051448391 @default.
• W2996344901 hasAuthorship W2996344901A5078711649 @default.
• W2996344901 hasAuthorship W2996344901A5085848346 @default.
• W2996344901 hasConcept C104317684 @default.
• W2996344901 hasConcept C119857082 @default.
• W2996344901 hasConcept C126255220 @default.
• W2996344901 hasConcept C147764199 @default.
• W2996344901 hasConcept C153083717 @default.
• W2996344901 hasConcept C154945302 @default.
• W2996344901 hasConcept C185592680 @default.
• W2996344901 hasConcept C199360897 @default.
• W2996344901 hasConcept C2776330181 @default.
• W2996344901 hasConcept C2984842247 @default.
• W2996344901 hasConcept C33923547 @default.
• W2996344901 hasConcept C37736160 @default.
• W2996344901 hasConcept C41008148 @default.
• W2996344901 hasConcept C50644808 @default.
• W2996344901 hasConcept C51632099 @default.
• W2996344901 hasConcept C55493867 @default.
• W2996344901 hasConcept C63479239 @default.
• W2996344901 hasConceptScore W2996344901C104317684 @default.
• W2996344901 hasConceptScore W2996344901C119857082 @default.
• W2996344901 hasConceptScore W2996344901C126255220 @default.
• W2996344901 hasConceptScore W2996344901C147764199 @default.
• W2996344901 hasConceptScore W2996344901C153083717 @default.
• W2996344901 hasConceptScore W2996344901C154945302 @default.
• W2996344901 hasConceptScore W2996344901C185592680 @default.
• W2996344901 hasConceptScore W2996344901C199360897 @default.
• W2996344901 hasConceptScore W2996344901C2776330181 @default.
• W2996344901 hasConceptScore W2996344901C2984842247 @default.

• next