Matches in SemOpenAlex for { <https://semopenalex.org/work/W3033415980> ?p ?o ?g. }
Showing items 1 to 78 of
78
with 100 items per page.
- W3033415980 abstract "Convolutional neural networks or standard CNNs (StdCNNs) are translation-equivariant models that achieve translation invariance when trained on data augmented with sufficient translations. Recent work on equivariant models for a given group of transformations (e.g., rotations) has lead to group-equivariant convolutional neural networks (GCNNs). GCNNs trained on data augmented with sufficient rotations achieve rotation invariance. Recent work by authors arXiv:2002.11318 studies a trade-off between invariance and robustness to adversarial attacks. In another related work arXiv:2005.08632, given any model and any input-dependent attack that satisfies a certain spectral property, the authors propose a universalization technique called SVD-Universal to produce a universal adversarial perturbation by looking at very few test examples. In this paper, we study the effectiveness of SVD-Universal on GCNNs as they gain rotation invariance through higher degree of training augmentation. We empirically observe that as GCNNs gain rotation invariance through training augmented with larger rotations, the fooling rate of SVD-Universal gets better. To understand this phenomenon, we introduce universal invariant directions and study their relation to the universal adversarial direction produced by SVD-Universal." @default.
- W3033415980 created "2020-06-12" @default.
- W3033415980 creator A5018784541 @default.
- W3033415980 creator A5029414285 @default.
- W3033415980 creator A5031047596 @default.
- W3033415980 date "2020-06-08" @default.
- W3033415980 modified "2023-09-25" @default.
- W3033415980 title "On Universalized Adversarial and Invariant Perturbations" @default.
- W3033415980 cites W2194775991 @default.
- W3033415980 cites W2259303769 @default.
- W3033415980 cites W2279221249 @default.
- W3033415980 cites W2543927648 @default.
- W3033415980 cites W2569680626 @default.
- W3033415980 cites W2618098489 @default.
- W3033415980 cites W2620411679 @default.
- W3033415980 cites W2736899637 @default.
- W3033415980 cites W2785887350 @default.
- W3033415980 cites W2788198349 @default.
- W3033415980 cites W2810443692 @default.
- W3033415980 cites W2885183727 @default.
- W3033415980 cites W2951320766 @default.
- W3033415980 cites W2963504932 @default.
- W3033415980 cites W2963564809 @default.
- W3033415980 cites W2963809642 @default.
- W3033415980 cites W3025116310 @default.
- W3033415980 cites W3025311475 @default.
- W3033415980 cites W3138516063 @default.
- W3033415980 doi "https://doi.org/10.48550/arxiv.2006.04449" @default.
- W3033415980 hasPublicationYear "2020" @default.
- W3033415980 type Work @default.
- W3033415980 sameAs 3033415980 @default.
- W3033415980 citedByCount "0" @default.
- W3033415980 crossrefType "posted-content" @default.
- W3033415980 hasAuthorship W3033415980A5018784541 @default.
- W3033415980 hasAuthorship W3033415980A5029414285 @default.
- W3033415980 hasAuthorship W3033415980A5031047596 @default.
- W3033415980 hasBestOaLocation W30334159801 @default.
- W3033415980 hasConcept C11413529 @default.
- W3033415980 hasConcept C154945302 @default.
- W3033415980 hasConcept C171036898 @default.
- W3033415980 hasConcept C190470478 @default.
- W3033415980 hasConcept C202444582 @default.
- W3033415980 hasConcept C22789450 @default.
- W3033415980 hasConcept C33098168 @default.
- W3033415980 hasConcept C33923547 @default.
- W3033415980 hasConcept C37736160 @default.
- W3033415980 hasConcept C37914503 @default.
- W3033415980 hasConcept C41008148 @default.
- W3033415980 hasConcept C81363708 @default.
- W3033415980 hasConceptScore W3033415980C11413529 @default.
- W3033415980 hasConceptScore W3033415980C154945302 @default.
- W3033415980 hasConceptScore W3033415980C171036898 @default.
- W3033415980 hasConceptScore W3033415980C190470478 @default.
- W3033415980 hasConceptScore W3033415980C202444582 @default.
- W3033415980 hasConceptScore W3033415980C22789450 @default.
- W3033415980 hasConceptScore W3033415980C33098168 @default.
- W3033415980 hasConceptScore W3033415980C33923547 @default.
- W3033415980 hasConceptScore W3033415980C37736160 @default.
- W3033415980 hasConceptScore W3033415980C37914503 @default.
- W3033415980 hasConceptScore W3033415980C41008148 @default.
- W3033415980 hasConceptScore W3033415980C81363708 @default.
- W3033415980 hasLocation W30334159801 @default.
- W3033415980 hasOpenAccess W3033415980 @default.
- W3033415980 hasPrimaryLocation W30334159801 @default.
- W3033415980 hasRelatedWork W2027233917 @default.
- W3033415980 hasRelatedWork W2374789424 @default.
- W3033415980 hasRelatedWork W2393434063 @default.
- W3033415980 hasRelatedWork W2793523722 @default.
- W3033415980 hasRelatedWork W2946727698 @default.
- W3033415980 hasRelatedWork W2949548663 @default.
- W3033415980 hasRelatedWork W3119542136 @default.
- W3033415980 hasRelatedWork W3194550388 @default.
- W3033415980 hasRelatedWork W3198296966 @default.
- W3033415980 hasRelatedWork W4287760661 @default.
- W3033415980 isParatext "false" @default.
- W3033415980 isRetracted "false" @default.
- W3033415980 magId "3033415980" @default.
- W3033415980 workType "article" @default.