Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386075662> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W4386075662 abstract "Deep Neural Networks show superior performance in various tasks but are vulnerable to adversarial attacks. Most defense techniques are devoted to the adversarial training strategies, however, it is difficult to achieve satisfactory robust performance only with traditional adversarial training. We mainly attribute it to that aggressive perturbations which lead to the loss increment can always be found via gradient ascent in white-box setting. Although some noises can be involved to prevent attacks from deriving precise gradients on inputs, there exist trade-offs between the defense capability and natural generalization. Taking advantage of the properties of random projection, we propose to replace part of convolutional filters with random projection filters, and theoretically explore the geometric representation preservation of proposed synthesized filters via Johnson-Lindenstrauss lemma. We conduct sufficient evaluation on multiple networks and datasets. The experimental results showcase the superiority of proposed random projection filters to state-of-the-art baselines. The code is available on GitHub." @default.
- W4386075662 created "2023-08-23" @default.
- W4386075662 creator A5001529504 @default.
- W4386075662 creator A5084617283 @default.
- W4386075662 date "2023-06-01" @default.
- W4386075662 modified "2023-09-30" @default.
- W4386075662 title "Adversarial Robustness via Random Projection Filters" @default.
- W4386075662 cites W2089497633 @default.
- W4386075662 cites W2194775991 @default.
- W4386075662 cites W2243397390 @default.
- W4386075662 cites W2774644650 @default.
- W4386075662 cites W2963485691 @default.
- W4386075662 cites W2963542245 @default.
- W4386075662 cites W2963857521 @default.
- W4386075662 cites W3034802054 @default.
- W4386075662 cites W3171288285 @default.
- W4386075662 cites W3179916154 @default.
- W4386075662 cites W4224285564 @default.
- W4386075662 cites W4312599584 @default.
- W4386075662 cites W4312692109 @default.
- W4386075662 cites W4313140285 @default.
- W4386075662 doi "https://doi.org/10.1109/cvpr52729.2023.00397" @default.
- W4386075662 hasPublicationYear "2023" @default.
- W4386075662 type Work @default.
- W4386075662 citedByCount "0" @default.
- W4386075662 crossrefType "proceedings-article" @default.
- W4386075662 hasAuthorship W4386075662A5001529504 @default.
- W4386075662 hasAuthorship W4386075662A5084617283 @default.
- W4386075662 hasConcept C104317684 @default.
- W4386075662 hasConcept C11413529 @default.
- W4386075662 hasConcept C119857082 @default.
- W4386075662 hasConcept C134306372 @default.
- W4386075662 hasConcept C154945302 @default.
- W4386075662 hasConcept C177148314 @default.
- W4386075662 hasConcept C17744445 @default.
- W4386075662 hasConcept C185592680 @default.
- W4386075662 hasConcept C18903297 @default.
- W4386075662 hasConcept C199539241 @default.
- W4386075662 hasConcept C2776359362 @default.
- W4386075662 hasConcept C2777036070 @default.
- W4386075662 hasConcept C2777759810 @default.
- W4386075662 hasConcept C2984842247 @default.
- W4386075662 hasConcept C33923547 @default.
- W4386075662 hasConcept C37736160 @default.
- W4386075662 hasConcept C41008148 @default.
- W4386075662 hasConcept C46757340 @default.
- W4386075662 hasConcept C50644808 @default.
- W4386075662 hasConcept C55493867 @default.
- W4386075662 hasConcept C57493831 @default.
- W4386075662 hasConcept C63479239 @default.
- W4386075662 hasConcept C81363708 @default.
- W4386075662 hasConcept C86803240 @default.
- W4386075662 hasConcept C94625758 @default.
- W4386075662 hasConceptScore W4386075662C104317684 @default.
- W4386075662 hasConceptScore W4386075662C11413529 @default.
- W4386075662 hasConceptScore W4386075662C119857082 @default.
- W4386075662 hasConceptScore W4386075662C134306372 @default.
- W4386075662 hasConceptScore W4386075662C154945302 @default.
- W4386075662 hasConceptScore W4386075662C177148314 @default.
- W4386075662 hasConceptScore W4386075662C17744445 @default.
- W4386075662 hasConceptScore W4386075662C185592680 @default.
- W4386075662 hasConceptScore W4386075662C18903297 @default.
- W4386075662 hasConceptScore W4386075662C199539241 @default.
- W4386075662 hasConceptScore W4386075662C2776359362 @default.
- W4386075662 hasConceptScore W4386075662C2777036070 @default.
- W4386075662 hasConceptScore W4386075662C2777759810 @default.
- W4386075662 hasConceptScore W4386075662C2984842247 @default.
- W4386075662 hasConceptScore W4386075662C33923547 @default.
- W4386075662 hasConceptScore W4386075662C37736160 @default.
- W4386075662 hasConceptScore W4386075662C41008148 @default.
- W4386075662 hasConceptScore W4386075662C46757340 @default.
- W4386075662 hasConceptScore W4386075662C50644808 @default.
- W4386075662 hasConceptScore W4386075662C55493867 @default.
- W4386075662 hasConceptScore W4386075662C57493831 @default.
- W4386075662 hasConceptScore W4386075662C63479239 @default.
- W4386075662 hasConceptScore W4386075662C81363708 @default.
- W4386075662 hasConceptScore W4386075662C86803240 @default.
- W4386075662 hasConceptScore W4386075662C94625758 @default.
- W4386075662 hasFunder F4320334704 @default.
- W4386075662 hasLocation W43860756621 @default.
- W4386075662 hasOpenAccess W4386075662 @default.
- W4386075662 hasPrimaryLocation W43860756621 @default.
- W4386075662 hasRelatedWork W2950183588 @default.
- W4386075662 hasRelatedWork W3072584680 @default.
- W4386075662 hasRelatedWork W3094843325 @default.
- W4386075662 hasRelatedWork W3207178610 @default.
- W4386075662 hasRelatedWork W4288362200 @default.
- W4386075662 hasRelatedWork W4311734044 @default.
- W4386075662 hasRelatedWork W4322759769 @default.
- W4386075662 hasRelatedWork W4365420856 @default.
- W4386075662 hasRelatedWork W4382763378 @default.
- W4386075662 hasRelatedWork W4385415357 @default.
- W4386075662 isParatext "false" @default.
- W4386075662 isRetracted "false" @default.
- W4386075662 workType "article" @default.