Matches in SemOpenAlex for { <https://semopenalex.org/work/W3035527730> ?p ?o ?g. }
- W3035527730 abstract "This paper studies how encouraging semantically-aligned features during deep neural network training can increase network robustness. Recent works observed that Adversarial Training leads to robust models, whose learnt features appear to correlate with human perception. Inspired by this connection from robustness to semantics, we study the complementary connection: from semantics to robustness. To do so, we provide a tight robustness certificate for distance-based classification models (clustering-based classifiers), which we leverage to propose ClusTR (Clustering Training for Robustness), a clustering-based and adversary-free training framework to learn robust models. Interestingly, ClusTR outperforms adversarially-trained networks by up to 4% under strong PGD attacks. Moreover, it can be equipped with simple and fast adversarial training to improve the current state-of-the-art in robustness by 16%-29% on CIFAR10, SVHN, and CIFAR100." @default.
- W3035527730 created "2020-06-19" @default.
- W3035527730 creator A5013512298 @default.
- W3035527730 creator A5016028605 @default.
- W3035527730 creator A5024763828 @default.
- W3035527730 creator A5026143507 @default.
- W3035527730 creator A5091298335 @default.
- W3035527730 creator A5091670711 @default.
- W3035527730 date "2020-06-13" @default.
- W3035527730 modified "2023-09-28" @default.
- W3035527730 title "ClusTR: Clustering Training for Robustness." @default.
- W3035527730 cites W1522301498 @default.
- W3035527730 cites W1757796397 @default.
- W3035527730 cites W2073459066 @default.
- W3035527730 cites W2106053110 @default.
- W3035527730 cites W2117154949 @default.
- W3035527730 cites W2123024445 @default.
- W3035527730 cites W2129156852 @default.
- W3035527730 cites W2145339207 @default.
- W3035527730 cites W2160815625 @default.
- W3035527730 cites W2194775991 @default.
- W3035527730 cites W2335728318 @default.
- W3035527730 cites W2610190180 @default.
- W3035527730 cites W2626778328 @default.
- W3035527730 cites W2786118190 @default.
- W3035527730 cites W2787708942 @default.
- W3035527730 cites W2791953061 @default.
- W3035527730 cites W2799032899 @default.
- W3035527730 cites W2897355816 @default.
- W3035527730 cites W2912237282 @default.
- W3035527730 cites W2913266441 @default.
- W3035527730 cites W2942630857 @default.
- W3035527730 cites W2950133940 @default.
- W3035527730 cites W2962872506 @default.
- W3035527730 cites W2962887033 @default.
- W3035527730 cites W2963026800 @default.
- W3035527730 cites W2963158386 @default.
- W3035527730 cites W2963775347 @default.
- W3035527730 cites W2964082701 @default.
- W3035527730 cites W2964153729 @default.
- W3035527730 cites W2964253222 @default.
- W3035527730 cites W2964308564 @default.
- W3035527730 cites W2969656782 @default.
- W3035527730 cites W2970317235 @default.
- W3035527730 cites W2970680991 @default.
- W3035527730 cites W2970971581 @default.
- W3035527730 cites W2976501124 @default.
- W3035527730 cites W2998835636 @default.
- W3035527730 cites W3103340107 @default.
- W3035527730 cites W3118608800 @default.
- W3035527730 cites W93016980 @default.
- W3035527730 hasPublicationYear "2020" @default.
- W3035527730 type Work @default.
- W3035527730 sameAs 3035527730 @default.
- W3035527730 citedByCount "2" @default.
- W3035527730 countsByYear W30355277302020 @default.
- W3035527730 countsByYear W30355277302021 @default.
- W3035527730 crossrefType "posted-content" @default.
- W3035527730 hasAuthorship W3035527730A5013512298 @default.
- W3035527730 hasAuthorship W3035527730A5016028605 @default.
- W3035527730 hasAuthorship W3035527730A5024763828 @default.
- W3035527730 hasAuthorship W3035527730A5026143507 @default.
- W3035527730 hasAuthorship W3035527730A5091298335 @default.
- W3035527730 hasAuthorship W3035527730A5091670711 @default.
- W3035527730 hasConcept C104317684 @default.
- W3035527730 hasConcept C119857082 @default.
- W3035527730 hasConcept C124101348 @default.
- W3035527730 hasConcept C153083717 @default.
- W3035527730 hasConcept C154945302 @default.
- W3035527730 hasConcept C185592680 @default.
- W3035527730 hasConcept C37736160 @default.
- W3035527730 hasConcept C41008148 @default.
- W3035527730 hasConcept C50644808 @default.
- W3035527730 hasConcept C51632099 @default.
- W3035527730 hasConcept C55493867 @default.
- W3035527730 hasConcept C63479239 @default.
- W3035527730 hasConcept C73555534 @default.
- W3035527730 hasConceptScore W3035527730C104317684 @default.
- W3035527730 hasConceptScore W3035527730C119857082 @default.
- W3035527730 hasConceptScore W3035527730C124101348 @default.
- W3035527730 hasConceptScore W3035527730C153083717 @default.
- W3035527730 hasConceptScore W3035527730C154945302 @default.
- W3035527730 hasConceptScore W3035527730C185592680 @default.
- W3035527730 hasConceptScore W3035527730C37736160 @default.
- W3035527730 hasConceptScore W3035527730C41008148 @default.
- W3035527730 hasConceptScore W3035527730C50644808 @default.
- W3035527730 hasConceptScore W3035527730C51632099 @default.
- W3035527730 hasConceptScore W3035527730C55493867 @default.
- W3035527730 hasConceptScore W3035527730C63479239 @default.
- W3035527730 hasConceptScore W3035527730C73555534 @default.
- W3035527730 hasLocation W30355277301 @default.
- W3035527730 hasOpenAccess W3035527730 @default.
- W3035527730 hasPrimaryLocation W30355277301 @default.
- W3035527730 hasRelatedWork W2948398419 @default.
- W3035527730 hasRelatedWork W2962872506 @default.
- W3035527730 hasRelatedWork W2970088379 @default.
- W3035527730 hasRelatedWork W3004298045 @default.
- W3035527730 hasRelatedWork W3005295611 @default.
- W3035527730 hasRelatedWork W3018992239 @default.
- W3035527730 hasRelatedWork W3034190247 @default.