Matches in SemOpenAlex for { <https://semopenalex.org/work/W2973934518> ?p ?o ?g. }
- W2973934518 abstract "Adversarial training has been demonstrated as one of the most effective methods for training robust models to defend against adversarial examples. However, adversarially trained models often lack adversarially robust generalization on unseen testing data. Recent works show that adversarially trained models are more biased towards global structure features. Instead, in this work, we would like to investigate the relationship between the generalization of adversarial training and the robust local features, as the robust local features generalize well for unseen shape variation. To learn the robust local features, we develop a Random Block Shuffle (RBS) transformation to break up the global structure features on normal adversarial examples. We continue to propose a new approach called Robust Local Features for Adversarial Training (RLFAT), which first learns the robust local features by adversarial training on the RBS-transformed adversarial examples, and then transfers the robust local features into the training of normal adversarial examples. To demonstrate the generality of our argument, we implement RLFAT in currently state-of-the-art adversarial training frameworks. Extensive experiments on STL-10, CIFAR-10 and CIFAR-100 show that RLFAT significantly improves both the adversarially robust generalization and the standard generalization of adversarial training. Additionally, we demonstrate that our models capture more local features of the object on the images, aligning better with human perception." @default.
- W2973934518 created "2019-09-26" @default.
- W2973934518 creator A5033526822 @default.
- W2973934518 creator A5043211393 @default.
- W2973934518 creator A5055723755 @default.
- W2973934518 creator A5079317086 @default.
- W2973934518 creator A5083155503 @default.
- W2973934518 date "2019-09-23" @default.
- W2973934518 modified "2023-09-23" @default.
- W2973934518 title "Robust Local Features for Improving the Generalization of Adversarial Training" @default.
- W2973934518 cites W2118858186 @default.
- W2973934518 cites W2125186487 @default.
- W2973934518 cites W2163605009 @default.
- W2973934518 cites W2184045248 @default.
- W2973934518 cites W2570685808 @default.
- W2973934518 cites W2603766943 @default.
- W2973934518 cites W2626639386 @default.
- W2973934518 cites W2786118190 @default.
- W2973934518 cites W2902617128 @default.
- W2973934518 cites W2909133815 @default.
- W2973934518 cites W2937975254 @default.
- W2973934518 cites W2942630857 @default.
- W2973934518 cites W2945359720 @default.
- W2973934518 cites W2947294642 @default.
- W2973934518 cites W2962729158 @default.
- W2973934518 cites W2962972504 @default.
- W2973934518 cites W2963143631 @default.
- W2973934518 cites W2963207607 @default.
- W2973934518 cites W2963361074 @default.
- W2973934518 cites W2963564844 @default.
- W2973934518 cites W2963597678 @default.
- W2973934518 cites W2963857521 @default.
- W2973934518 cites W2963894448 @default.
- W2973934518 cites W2963936085 @default.
- W2973934518 cites W2964137095 @default.
- W2973934518 cites W2964153729 @default.
- W2973934518 cites W2964224652 @default.
- W2973934518 cites W2964253222 @default.
- W2973934518 cites W2996140774 @default.
- W2973934518 cites W3092696781 @default.
- W2973934518 cites W3118608800 @default.
- W2973934518 hasPublicationYear "2019" @default.
- W2973934518 type Work @default.
- W2973934518 sameAs 2973934518 @default.
- W2973934518 citedByCount "1" @default.
- W2973934518 countsByYear W29739345182021 @default.
- W2973934518 crossrefType "posted-content" @default.
- W2973934518 hasAuthorship W2973934518A5033526822 @default.
- W2973934518 hasAuthorship W2973934518A5043211393 @default.
- W2973934518 hasAuthorship W2973934518A5055723755 @default.
- W2973934518 hasAuthorship W2973934518A5079317086 @default.
- W2973934518 hasAuthorship W2973934518A5083155503 @default.
- W2973934518 hasConcept C119857082 @default.
- W2973934518 hasConcept C134306372 @default.
- W2973934518 hasConcept C153294291 @default.
- W2973934518 hasConcept C154945302 @default.
- W2973934518 hasConcept C15744967 @default.
- W2973934518 hasConcept C177148314 @default.
- W2973934518 hasConcept C185592680 @default.
- W2973934518 hasConcept C205649164 @default.
- W2973934518 hasConcept C2524010 @default.
- W2973934518 hasConcept C2777210771 @default.
- W2973934518 hasConcept C2777211547 @default.
- W2973934518 hasConcept C2780767217 @default.
- W2973934518 hasConcept C2781238097 @default.
- W2973934518 hasConcept C33923547 @default.
- W2973934518 hasConcept C37736160 @default.
- W2973934518 hasConcept C41008148 @default.
- W2973934518 hasConcept C542102704 @default.
- W2973934518 hasConcept C55493867 @default.
- W2973934518 hasConcept C98184364 @default.
- W2973934518 hasConceptScore W2973934518C119857082 @default.
- W2973934518 hasConceptScore W2973934518C134306372 @default.
- W2973934518 hasConceptScore W2973934518C153294291 @default.
- W2973934518 hasConceptScore W2973934518C154945302 @default.
- W2973934518 hasConceptScore W2973934518C15744967 @default.
- W2973934518 hasConceptScore W2973934518C177148314 @default.
- W2973934518 hasConceptScore W2973934518C185592680 @default.
- W2973934518 hasConceptScore W2973934518C205649164 @default.
- W2973934518 hasConceptScore W2973934518C2524010 @default.
- W2973934518 hasConceptScore W2973934518C2777210771 @default.
- W2973934518 hasConceptScore W2973934518C2777211547 @default.
- W2973934518 hasConceptScore W2973934518C2780767217 @default.
- W2973934518 hasConceptScore W2973934518C2781238097 @default.
- W2973934518 hasConceptScore W2973934518C33923547 @default.
- W2973934518 hasConceptScore W2973934518C37736160 @default.
- W2973934518 hasConceptScore W2973934518C41008148 @default.
- W2973934518 hasConceptScore W2973934518C542102704 @default.
- W2973934518 hasConceptScore W2973934518C55493867 @default.
- W2973934518 hasConceptScore W2973934518C98184364 @default.
- W2973934518 hasOpenAccess W2973934518 @default.
- W2973934518 hasRelatedWork W2245869277 @default.
- W2973934518 hasRelatedWork W2884169041 @default.
- W2973934518 hasRelatedWork W2950942529 @default.
- W2973934518 hasRelatedWork W2963501948 @default.
- W2973934518 hasRelatedWork W2963870211 @default.
- W2973934518 hasRelatedWork W2982324875 @default.
- W2973934518 hasRelatedWork W2995645057 @default.
- W2973934518 hasRelatedWork W2996687884 @default.
- W2973934518 hasRelatedWork W3007712033 @default.