Matches in SemOpenAlex for { <https://semopenalex.org/work/W2970049488> ?p ?o ?g. }
Showing items 1 to 99 of
99
with 100 items per page.
- W2970049488 endingPage "13833" @default.
- W2970049488 startingPage "13824" @default.
- W2970049488 abstract "Adversarial training is an effective methodology for training deep neural networks that are robust against adversarial, norm-bounded perturbations. However, the computational cost of adversarial training grows prohibitively as the size of the model and number of input dimensions increase. Further, training against less expensive and therefore weaker adversaries produces models that are robust against weak attacks but break down under attacks that are stronger. This is often attributed to the phenomenon of gradient obfuscation; such models have a highly non-linear loss surface in the vicinity of training examples, making it hard for gradient-based attacks to succeed even though adversarial examples still exist. In this work, we introduce a novel regularizer that encourages the loss to behave linearly in the vicinity of the training data, thereby penalizing gradient obfuscation while encouraging robustness. We show via extensive experiments on CIFAR-10 and ImageNet, that models trained with our regularizer avoid gradient obfuscation and can be trained significantly faster than adversarial training. Using this regularizer, we exceed current state of the art and achieve 47% adversarial accuracy for ImageNet with L-infinity norm adversarial perturbations of radius 4/255 under an untargeted, strong, white-box attack. Additionally, we match state of the art results for CIFAR-10 at 8/255." @default.
- W2970049488 created "2019-09-05" @default.
- W2970049488 creator A5013834379 @default.
- W2970049488 creator A5036322364 @default.
- W2970049488 creator A5046229829 @default.
- W2970049488 creator A5050954769 @default.
- W2970049488 creator A5057393538 @default.
- W2970049488 creator A5059632022 @default.
- W2970049488 creator A5059925606 @default.
- W2970049488 creator A5076040849 @default.
- W2970049488 creator A5076073500 @default.
- W2970049488 date "2019-07-04" @default.
- W2970049488 modified "2023-09-24" @default.
- W2970049488 title "Adversarial Robustness through Local Linearization" @default.
- W2970049488 hasPublicationYear "2019" @default.
- W2970049488 type Work @default.
- W2970049488 sameAs 2970049488 @default.
- W2970049488 citedByCount "116" @default.
- W2970049488 countsByYear W29700494882018 @default.
- W2970049488 countsByYear W29700494882019 @default.
- W2970049488 countsByYear W29700494882020 @default.
- W2970049488 countsByYear W29700494882021 @default.
- W2970049488 countsByYear W29700494882022 @default.
- W2970049488 crossrefType "proceedings-article" @default.
- W2970049488 hasAuthorship W2970049488A5013834379 @default.
- W2970049488 hasAuthorship W2970049488A5036322364 @default.
- W2970049488 hasAuthorship W2970049488A5046229829 @default.
- W2970049488 hasAuthorship W2970049488A5050954769 @default.
- W2970049488 hasAuthorship W2970049488A5057393538 @default.
- W2970049488 hasAuthorship W2970049488A5059632022 @default.
- W2970049488 hasAuthorship W2970049488A5059925606 @default.
- W2970049488 hasAuthorship W2970049488A5076040849 @default.
- W2970049488 hasAuthorship W2970049488A5076073500 @default.
- W2970049488 hasConcept C104317684 @default.
- W2970049488 hasConcept C108583219 @default.
- W2970049488 hasConcept C11413529 @default.
- W2970049488 hasConcept C119857082 @default.
- W2970049488 hasConcept C126255220 @default.
- W2970049488 hasConcept C134306372 @default.
- W2970049488 hasConcept C154945302 @default.
- W2970049488 hasConcept C17744445 @default.
- W2970049488 hasConcept C185592680 @default.
- W2970049488 hasConcept C191795146 @default.
- W2970049488 hasConcept C199539241 @default.
- W2970049488 hasConcept C2984842247 @default.
- W2970049488 hasConcept C33923547 @default.
- W2970049488 hasConcept C34388435 @default.
- W2970049488 hasConcept C37736160 @default.
- W2970049488 hasConcept C41008148 @default.
- W2970049488 hasConcept C55493867 @default.
- W2970049488 hasConcept C63479239 @default.
- W2970049488 hasConceptScore W2970049488C104317684 @default.
- W2970049488 hasConceptScore W2970049488C108583219 @default.
- W2970049488 hasConceptScore W2970049488C11413529 @default.
- W2970049488 hasConceptScore W2970049488C119857082 @default.
- W2970049488 hasConceptScore W2970049488C126255220 @default.
- W2970049488 hasConceptScore W2970049488C134306372 @default.
- W2970049488 hasConceptScore W2970049488C154945302 @default.
- W2970049488 hasConceptScore W2970049488C17744445 @default.
- W2970049488 hasConceptScore W2970049488C185592680 @default.
- W2970049488 hasConceptScore W2970049488C191795146 @default.
- W2970049488 hasConceptScore W2970049488C199539241 @default.
- W2970049488 hasConceptScore W2970049488C2984842247 @default.
- W2970049488 hasConceptScore W2970049488C33923547 @default.
- W2970049488 hasConceptScore W2970049488C34388435 @default.
- W2970049488 hasConceptScore W2970049488C37736160 @default.
- W2970049488 hasConceptScore W2970049488C41008148 @default.
- W2970049488 hasConceptScore W2970049488C55493867 @default.
- W2970049488 hasConceptScore W2970049488C63479239 @default.
- W2970049488 hasLocation W29700494881 @default.
- W2970049488 hasOpenAccess W2970049488 @default.
- W2970049488 hasPrimaryLocation W29700494881 @default.
- W2970049488 hasRelatedWork W2194775991 @default.
- W2970049488 hasRelatedWork W2243397390 @default.
- W2970049488 hasRelatedWork W2791953061 @default.
- W2970049488 hasRelatedWork W2911634294 @default.
- W2970049488 hasRelatedWork W2912237282 @default.
- W2970049488 hasRelatedWork W2913266441 @default.
- W2970049488 hasRelatedWork W2962872506 @default.
- W2970049488 hasRelatedWork W2963026800 @default.
- W2970049488 hasRelatedWork W2963143631 @default.
- W2970049488 hasRelatedWork W2963207607 @default.
- W2970049488 hasRelatedWork W2963389226 @default.
- W2970049488 hasRelatedWork W2963857521 @default.
- W2970049488 hasRelatedWork W2964116600 @default.
- W2970049488 hasRelatedWork W2964153729 @default.
- W2970049488 hasRelatedWork W2964253222 @default.
- W2970049488 hasRelatedWork W2970317235 @default.
- W2970049488 hasRelatedWork W2970680991 @default.
- W2970049488 hasRelatedWork W2971316968 @default.
- W2970049488 hasRelatedWork W3034994123 @default.
- W2970049488 hasRelatedWork W3118608800 @default.
- W2970049488 hasVolume "32" @default.
- W2970049488 isParatext "false" @default.
- W2970049488 isRetracted "false" @default.
- W2970049488 magId "2970049488" @default.
- W2970049488 workType "article" @default.