Matches in SemOpenAlex for { <https://semopenalex.org/work/W4289550810> ?p ?o ?g. }
Showing items 1 to 51 of
51
with 100 items per page.
- W4289550810 abstract "Short answer: Yes, Long answer: No! Indeed, research on adversarial robustness has led to invaluable insights helping us understand and explore different aspects of the problem. Many attacks and defenses have been proposed over the last couple of years. The problem, however, remains largely unsolved and poorly understood. Here, I argue that the current formulation of the problem serves short term goals, and needs to be revised for us to achieve bigger gains. Specifically, the bound on perturbation has created a somewhat contrived setting and needs to be relaxed. This has misled us to focus on model classes that are not expressive enough to begin with. Instead, inspired by human vision and the fact that we rely more on robust features such as shape, vertices, and foreground objects than non-robust features such as texture, efforts should be steered towards looking for significantly different classes of models. Maybe instead of narrowing down on imperceptible adversarial perturbations, we should attack a more general problem which is finding architectures that are simultaneously robust to perceptible perturbations, geometric transformations (e.g. rotation, scaling), image distortions (lighting, blur), and more (e.g. occlusion, shadow). Only then we may be able to solve the problem of adversarial vulnerability." @default.
- W4289550810 created "2022-08-03" @default.
- W4289550810 creator A5068743986 @default.
- W4289550810 date "2022-07-31" @default.
- W4289550810 modified "2023-09-29" @default.
- W4289550810 title "Is current research on adversarial robustness addressing the right problem?" @default.
- W4289550810 doi "https://doi.org/10.48550/arxiv.2208.00539" @default.
- W4289550810 hasPublicationYear "2022" @default.
- W4289550810 type Work @default.
- W4289550810 citedByCount "0" @default.
- W4289550810 crossrefType "posted-content" @default.
- W4289550810 hasAuthorship W4289550810A5068743986 @default.
- W4289550810 hasBestOaLocation W42895508101 @default.
- W4289550810 hasConcept C104317684 @default.
- W4289550810 hasConcept C154945302 @default.
- W4289550810 hasConcept C185592680 @default.
- W4289550810 hasConcept C2524010 @default.
- W4289550810 hasConcept C31972630 @default.
- W4289550810 hasConcept C33923547 @default.
- W4289550810 hasConcept C37736160 @default.
- W4289550810 hasConcept C41008148 @default.
- W4289550810 hasConcept C55493867 @default.
- W4289550810 hasConcept C63479239 @default.
- W4289550810 hasConcept C99844830 @default.
- W4289550810 hasConceptScore W4289550810C104317684 @default.
- W4289550810 hasConceptScore W4289550810C154945302 @default.
- W4289550810 hasConceptScore W4289550810C185592680 @default.
- W4289550810 hasConceptScore W4289550810C2524010 @default.
- W4289550810 hasConceptScore W4289550810C31972630 @default.
- W4289550810 hasConceptScore W4289550810C33923547 @default.
- W4289550810 hasConceptScore W4289550810C37736160 @default.
- W4289550810 hasConceptScore W4289550810C41008148 @default.
- W4289550810 hasConceptScore W4289550810C55493867 @default.
- W4289550810 hasConceptScore W4289550810C63479239 @default.
- W4289550810 hasConceptScore W4289550810C99844830 @default.
- W4289550810 hasLocation W42895508101 @default.
- W4289550810 hasOpenAccess W4289550810 @default.
- W4289550810 hasPrimaryLocation W42895508101 @default.
- W4289550810 hasRelatedWork W10096406 @default.
- W4289550810 hasRelatedWork W10124904 @default.
- W4289550810 hasRelatedWork W11882198 @default.
- W4289550810 hasRelatedWork W13255732 @default.
- W4289550810 hasRelatedWork W14235181 @default.
- W4289550810 hasRelatedWork W2279739 @default.
- W4289550810 hasRelatedWork W4320778 @default.
- W4289550810 hasRelatedWork W4364146 @default.
- W4289550810 hasRelatedWork W5134648 @default.
- W4289550810 hasRelatedWork W9124666 @default.
- W4289550810 isParatext "false" @default.
- W4289550810 isRetracted "false" @default.
- W4289550810 workType "article" @default.