Matches in SemOpenAlex for { <https://semopenalex.org/work/W3109453310> ?p ?o ?g. }
- W3109453310 abstract "Previous adversarial training raises model robustness under the compromise of accuracy on natural data. In this paper, we reduce natural accuracy degradation. We use the model logits from one clean model to guide learning of another one robust model, taking into consideration that logits from the well trained clean model embed the most discriminative features of natural data, {it e.g.}, generalizable classifier boundary. Our solution is to constrain logits from the robust model that takes adversarial examples as input and makes it similar to those from the clean model fed with corresponding natural data. It lets the robust model inherit the classifier boundary of the clean model. Moreover, we observe such boundary guidance can not only preserve high natural accuracy but also benefit model robustness, which gives new insights and facilitates progress for the adversarial community. Finally, extensive experiments on CIFAR-10, CIFAR-100, and Tiny ImageNet testify to the effectiveness of our method. We achieve new state-of-the-art robustness on CIFAR-100 without additional real or synthetic data with auto-attack benchmark footnote{url{https://github.com/fra31/auto-attack}}. Our code is available at url{https://github.com/dvlab-research/LBGAT}." @default.
- W3109453310 created "2020-12-07" @default.
- W3109453310 creator A5016170160 @default.
- W3109453310 creator A5052856441 @default.
- W3109453310 creator A5055723755 @default.
- W3109453310 creator A5085253437 @default.
- W3109453310 date "2020-11-22" @default.
- W3109453310 modified "2023-10-14" @default.
- W3109453310 title "Learnable Boundary Guided Adversarial Training" @default.
- W3109453310 cites W1821462560 @default.
- W3109453310 cites W2108598243 @default.
- W3109453310 cites W2194775991 @default.
- W3109453310 cites W2243397390 @default.
- W3109453310 cites W2541993794 @default.
- W3109453310 cites W2607219512 @default.
- W3109453310 cites W2766966408 @default.
- W3109453310 cites W2774644650 @default.
- W3109453310 cites W2791953061 @default.
- W3109453310 cites W2810075754 @default.
- W3109453310 cites W2912237282 @default.
- W3109453310 cites W2946089113 @default.
- W3109453310 cites W2948558869 @default.
- W3109453310 cites W2962729158 @default.
- W3109453310 cites W2962835968 @default.
- W3109453310 cites W2962847335 @default.
- W3109453310 cites W2963140444 @default.
- W3109453310 cites W2963143631 @default.
- W3109453310 cites W2963207607 @default.
- W3109453310 cites W2963446712 @default.
- W3109453310 cites W2963516603 @default.
- W3109453310 cites W2963542245 @default.
- W3109453310 cites W2963744840 @default.
- W3109453310 cites W2963857521 @default.
- W3109453310 cites W2963918968 @default.
- W3109453310 cites W2963920068 @default.
- W3109453310 cites W2964082701 @default.
- W3109453310 cites W2964137095 @default.
- W3109453310 cites W2964153729 @default.
- W3109453310 cites W2964224652 @default.
- W3109453310 cites W2964253222 @default.
- W3109453310 cites W2964449891 @default.
- W3109453310 cites W2970049488 @default.
- W3109453310 cites W2970088379 @default.
- W3109453310 cites W2970316625 @default.
- W3109453310 cites W2970317235 @default.
- W3109453310 cites W2982242214 @default.
- W3109453310 cites W2982648698 @default.
- W3109453310 cites W2988089665 @default.
- W3109453310 cites W2994749257 @default.
- W3109453310 cites W2994897359 @default.
- W3109453310 cites W2995554640 @default.
- W3109453310 cites W2995607862 @default.
- W3109453310 cites W2996564870 @default.
- W3109453310 cites W2996687884 @default.
- W3109453310 cites W3004298045 @default.
- W3109453310 cites W3012575288 @default.
- W3109453310 cites W3034994123 @default.
- W3109453310 cites W3035032188 @default.
- W3109453310 cites W3035106315 @default.
- W3109453310 cites W3035345420 @default.
- W3109453310 cites W3035743198 @default.
- W3109453310 cites W3089774673 @default.
- W3109453310 cites W3099502074 @default.
- W3109453310 cites W3107235539 @default.
- W3109453310 cites W3136604105 @default.
- W3109453310 cites W3177196641 @default.
- W3109453310 doi "https://doi.org/10.48550/arxiv.2011.11164" @default.
- W3109453310 hasPublicationYear "2020" @default.
- W3109453310 type Work @default.
- W3109453310 sameAs 3109453310 @default.
- W3109453310 citedByCount "9" @default.
- W3109453310 countsByYear W31094533102021 @default.
- W3109453310 crossrefType "posted-content" @default.
- W3109453310 hasAuthorship W3109453310A5016170160 @default.
- W3109453310 hasAuthorship W3109453310A5052856441 @default.
- W3109453310 hasAuthorship W3109453310A5055723755 @default.
- W3109453310 hasAuthorship W3109453310A5085253437 @default.
- W3109453310 hasBestOaLocation W31094533101 @default.
- W3109453310 hasConcept C104317684 @default.
- W3109453310 hasConcept C119857082 @default.
- W3109453310 hasConcept C124101348 @default.
- W3109453310 hasConcept C154945302 @default.
- W3109453310 hasConcept C185592680 @default.
- W3109453310 hasConcept C37736160 @default.
- W3109453310 hasConcept C41008148 @default.
- W3109453310 hasConcept C42023084 @default.
- W3109453310 hasConcept C51632099 @default.
- W3109453310 hasConcept C55493867 @default.
- W3109453310 hasConcept C63479239 @default.
- W3109453310 hasConcept C95623464 @default.
- W3109453310 hasConcept C97931131 @default.
- W3109453310 hasConceptScore W3109453310C104317684 @default.
- W3109453310 hasConceptScore W3109453310C119857082 @default.
- W3109453310 hasConceptScore W3109453310C124101348 @default.
- W3109453310 hasConceptScore W3109453310C154945302 @default.
- W3109453310 hasConceptScore W3109453310C185592680 @default.
- W3109453310 hasConceptScore W3109453310C37736160 @default.
- W3109453310 hasConceptScore W3109453310C41008148 @default.
- W3109453310 hasConceptScore W3109453310C42023084 @default.
- W3109453310 hasConceptScore W3109453310C51632099 @default.