Matches in SemOpenAlex for { <https://semopenalex.org/work/W2899434523> ?p ?o ?g. }
- W2899434523 abstract "Adversarial examples are a pervasive phenomenon of machine learning models where seemingly imperceptible perturbations to the input lead to misclassifications for otherwise statistically accurate models. We propose a geometric framework, drawing on tools from the manifold reconstruction literature, to analyze the high-dimensional geometry of adversarial examples. In particular, we highlight the importance of codimension: for low-dimensional data manifolds embedded in high-dimensional space there are many directions off the manifold in which to construct adversarial examples. Adversarial examples are a natural consequence of learning a decision boundary that classifies the low-dimensional data manifold well, but classifies points near the manifold incorrectly. Using our geometric framework we prove (1) a tradeoff between robustness under different norms, (2) that adversarial training in balls around the data is sample inefficient, and (3) sufficient sampling conditions under which nearest neighbor classifiers and ball-based adversarial training are robust." @default.
- W2899434523 created "2018-11-09" @default.
- W2899434523 creator A5022522148 @default.
- W2899434523 creator A5076757561 @default.
- W2899434523 date "2018-09-27" @default.
- W2899434523 modified "2023-09-27" @default.
- W2899434523 title "On the Geometry of Adversarial Examples." @default.
- W2899434523 cites W1510073064 @default.
- W2899434523 cites W1673923490 @default.
- W2899434523 cites W1826955156 @default.
- W2899434523 cites W1945616565 @default.
- W2899434523 cites W1980267728 @default.
- W2899434523 cites W2016252883 @default.
- W2899434523 cites W2021414830 @default.
- W2899434523 cites W2076523440 @default.
- W2899434523 cites W2083485913 @default.
- W2899434523 cites W2083493892 @default.
- W2899434523 cites W2106209520 @default.
- W2899434523 cites W2119821739 @default.
- W2899434523 cites W2126439213 @default.
- W2899434523 cites W2292929658 @default.
- W2899434523 cites W2460937040 @default.
- W2899434523 cites W2506146024 @default.
- W2899434523 cites W2525778437 @default.
- W2899434523 cites W2594407953 @default.
- W2899434523 cites W2603766943 @default.
- W2899434523 cites W2619516334 @default.
- W2899434523 cites W2626166918 @default.
- W2899434523 cites W2760878839 @default.
- W2899434523 cites W2766462876 @default.
- W2899434523 cites W2767075075 @default.
- W2899434523 cites W2787708942 @default.
- W2899434523 cites W2799032899 @default.
- W2899434523 cites W2810611310 @default.
- W2899434523 cites W2885241258 @default.
- W2899434523 cites W2911742574 @default.
- W2899434523 cites W2952842465 @default.
- W2899434523 cites W2963070423 @default.
- W2899434523 cites W2963982496 @default.
- W2899434523 cites W2964121744 @default.
- W2899434523 cites W2964253222 @default.
- W2899434523 cites W582625930 @default.
- W2899434523 hasPublicationYear "2018" @default.
- W2899434523 type Work @default.
- W2899434523 sameAs 2899434523 @default.
- W2899434523 citedByCount "23" @default.
- W2899434523 countsByYear W28994345232018 @default.
- W2899434523 countsByYear W28994345232019 @default.
- W2899434523 countsByYear W28994345232020 @default.
- W2899434523 countsByYear W28994345232021 @default.
- W2899434523 crossrefType "posted-content" @default.
- W2899434523 hasAuthorship W2899434523A5022522148 @default.
- W2899434523 hasAuthorship W2899434523A5076757561 @default.
- W2899434523 hasConcept C104317684 @default.
- W2899434523 hasConcept C11413529 @default.
- W2899434523 hasConcept C122041747 @default.
- W2899434523 hasConcept C127413603 @default.
- W2899434523 hasConcept C151876577 @default.
- W2899434523 hasConcept C154945302 @default.
- W2899434523 hasConcept C185592680 @default.
- W2899434523 hasConcept C199360897 @default.
- W2899434523 hasConcept C202444582 @default.
- W2899434523 hasConcept C2524010 @default.
- W2899434523 hasConcept C2780801425 @default.
- W2899434523 hasConcept C33923547 @default.
- W2899434523 hasConcept C37736160 @default.
- W2899434523 hasConcept C41008148 @default.
- W2899434523 hasConcept C42023084 @default.
- W2899434523 hasConcept C529865628 @default.
- W2899434523 hasConcept C55493867 @default.
- W2899434523 hasConcept C63479239 @default.
- W2899434523 hasConcept C70518039 @default.
- W2899434523 hasConcept C78519656 @default.
- W2899434523 hasConcept C83979697 @default.
- W2899434523 hasConcept C95623464 @default.
- W2899434523 hasConceptScore W2899434523C104317684 @default.
- W2899434523 hasConceptScore W2899434523C11413529 @default.
- W2899434523 hasConceptScore W2899434523C122041747 @default.
- W2899434523 hasConceptScore W2899434523C127413603 @default.
- W2899434523 hasConceptScore W2899434523C151876577 @default.
- W2899434523 hasConceptScore W2899434523C154945302 @default.
- W2899434523 hasConceptScore W2899434523C185592680 @default.
- W2899434523 hasConceptScore W2899434523C199360897 @default.
- W2899434523 hasConceptScore W2899434523C202444582 @default.
- W2899434523 hasConceptScore W2899434523C2524010 @default.
- W2899434523 hasConceptScore W2899434523C2780801425 @default.
- W2899434523 hasConceptScore W2899434523C33923547 @default.
- W2899434523 hasConceptScore W2899434523C37736160 @default.
- W2899434523 hasConceptScore W2899434523C41008148 @default.
- W2899434523 hasConceptScore W2899434523C42023084 @default.
- W2899434523 hasConceptScore W2899434523C529865628 @default.
- W2899434523 hasConceptScore W2899434523C55493867 @default.
- W2899434523 hasConceptScore W2899434523C63479239 @default.
- W2899434523 hasConceptScore W2899434523C70518039 @default.
- W2899434523 hasConceptScore W2899434523C78519656 @default.
- W2899434523 hasConceptScore W2899434523C83979697 @default.
- W2899434523 hasConceptScore W2899434523C95623464 @default.
- W2899434523 hasLocation W28994345231 @default.
- W2899434523 hasOpenAccess W2899434523 @default.
- W2899434523 hasPrimaryLocation W28994345231 @default.