Matches in SemOpenAlex for { <https://semopenalex.org/work/W2996058201> ?p ?o ?g. }
Showing items 1 to 91 of
91
with 100 items per page.
- W2996058201 abstract "Ensuring robustness of Deep Neural Networks (DNNs) is crucial to their adoption in safety-critical applications such as self-driving cars, drones, and healthcare. Notably, DNNs are vulnerable to adversarial attacks in which small input perturbations can produce catastrophic misclassifications. In this work, we propose EMPIR, ensembles of quantized DNN models with different numerical precisions, as a new approach to increase robustness against adversarial attacks. EMPIR is based on the observation that quantized neural networks often demonstrate much higher robustness to adversarial attacks than full precision networks, but at the cost of a substantial loss in accuracy on the original (unperturbed) inputs. EMPIR overcomes this limitation to achieve the “best of both worlds”, i.e., the higher unperturbed accuracies of the full precision models combined with the higher robustness of the low precision models, by composing them in an ensemble. Further, as low precision DNN models have significantly lower computational and storage requirements than full precision models, EMPIR models only incur modest compute and memory overheads compared to a single full-precision model (<25% in our evaluations). We evaluate EMPIR across a suite of 3 different DNN tasks (MNIST, CIFAR-10 and ImageNet) and under 4 different adversarial attacks. Our results indicate that EMPIR boosts the average adversarial accuracies by 43.6%, 15.3% and 11.9% for the DNN models trained on the MNIST, CIFAR-10 and ImageNet datasets respectively, when compared to single full-precision models, without sacrificing accuracy on the unperturbed inputs." @default.
- W2996058201 created "2019-12-26" @default.
- W2996058201 creator A5009374923 @default.
- W2996058201 creator A5021284824 @default.
- W2996058201 creator A5065766721 @default.
- W2996058201 date "2020-04-30" @default.
- W2996058201 modified "2023-09-27" @default.
- W2996058201 title "EMPIR: Ensembles of Mixed Precision Deep Networks for Increased Robustness Against Adversarial Attacks" @default.
- W2996058201 cites W1534477342 @default.
- W2996058201 cites W2108598243 @default.
- W2996058201 cites W2112796928 @default.
- W2996058201 cites W2135293965 @default.
- W2996058201 cites W2180612164 @default.
- W2996058201 cites W2469490737 @default.
- W2996058201 cites W2524428287 @default.
- W2996058201 cites W2755956102 @default.
- W2996058201 cites W2884491628 @default.
- W2996058201 cites W2889797931 @default.
- W2996058201 cites W2912070915 @default.
- W2996058201 cites W2951055820 @default.
- W2996058201 cites W2962699575 @default.
- W2996058201 cites W2963114950 @default.
- W2996058201 cites W2963207607 @default.
- W2996058201 cites W2963249138 @default.
- W2996058201 cites W2963431851 @default.
- W2996058201 cites W2963542245 @default.
- W2996058201 cites W2963857521 @default.
- W2996058201 cites W2964082701 @default.
- W2996058201 cites W2964253222 @default.
- W2996058201 cites W3118608800 @default.
- W2996058201 hasPublicationYear "2020" @default.
- W2996058201 type Work @default.
- W2996058201 sameAs 2996058201 @default.
- W2996058201 citedByCount "7" @default.
- W2996058201 countsByYear W29960582012020 @default.
- W2996058201 countsByYear W29960582012021 @default.
- W2996058201 crossrefType "proceedings-article" @default.
- W2996058201 hasAuthorship W2996058201A5009374923 @default.
- W2996058201 hasAuthorship W2996058201A5021284824 @default.
- W2996058201 hasAuthorship W2996058201A5065766721 @default.
- W2996058201 hasConcept C104317684 @default.
- W2996058201 hasConcept C108583219 @default.
- W2996058201 hasConcept C119857082 @default.
- W2996058201 hasConcept C154945302 @default.
- W2996058201 hasConcept C185592680 @default.
- W2996058201 hasConcept C190502265 @default.
- W2996058201 hasConcept C2984842247 @default.
- W2996058201 hasConcept C37736160 @default.
- W2996058201 hasConcept C41008148 @default.
- W2996058201 hasConcept C50644808 @default.
- W2996058201 hasConcept C55493867 @default.
- W2996058201 hasConcept C63479239 @default.
- W2996058201 hasConceptScore W2996058201C104317684 @default.
- W2996058201 hasConceptScore W2996058201C108583219 @default.
- W2996058201 hasConceptScore W2996058201C119857082 @default.
- W2996058201 hasConceptScore W2996058201C154945302 @default.
- W2996058201 hasConceptScore W2996058201C185592680 @default.
- W2996058201 hasConceptScore W2996058201C190502265 @default.
- W2996058201 hasConceptScore W2996058201C2984842247 @default.
- W2996058201 hasConceptScore W2996058201C37736160 @default.
- W2996058201 hasConceptScore W2996058201C41008148 @default.
- W2996058201 hasConceptScore W2996058201C50644808 @default.
- W2996058201 hasConceptScore W2996058201C55493867 @default.
- W2996058201 hasConceptScore W2996058201C63479239 @default.
- W2996058201 hasLocation W29960582011 @default.
- W2996058201 hasOpenAccess W2996058201 @default.
- W2996058201 hasPrimaryLocation W29960582011 @default.
- W2996058201 hasRelatedWork W2163605009 @default.
- W2996058201 hasRelatedWork W2180612164 @default.
- W2996058201 hasRelatedWork W2194775991 @default.
- W2996058201 hasRelatedWork W2533641151 @default.
- W2996058201 hasRelatedWork W2746600820 @default.
- W2996058201 hasRelatedWork W2887603965 @default.
- W2996058201 hasRelatedWork W2912070915 @default.
- W2996058201 hasRelatedWork W2947947498 @default.
- W2996058201 hasRelatedWork W2953054173 @default.
- W2996058201 hasRelatedWork W2963207607 @default.
- W2996058201 hasRelatedWork W2963564844 @default.
- W2996058201 hasRelatedWork W2963857521 @default.
- W2996058201 hasRelatedWork W2964082701 @default.
- W2996058201 hasRelatedWork W2964153729 @default.
- W2996058201 hasRelatedWork W2964253222 @default.
- W2996058201 hasRelatedWork W2970317235 @default.
- W2996058201 hasRelatedWork W3018979822 @default.
- W2996058201 hasRelatedWork W3103340107 @default.
- W2996058201 hasRelatedWork W3118608800 @default.
- W2996058201 hasRelatedWork W3164903455 @default.
- W2996058201 isParatext "false" @default.
- W2996058201 isRetracted "false" @default.
- W2996058201 magId "2996058201" @default.
- W2996058201 workType "article" @default.