Matches in SemOpenAlex for { <https://semopenalex.org/work/W3184505055> ?p ?o ?g. }
Showing items 1 to 93 of
93
with 100 items per page.
- W3184505055 abstract "The susceptibility of deep neural networks to untrustworthy predictions, including out-of-distribution (OOD) data and adversarial examples, still prevent their widespread use in safety-critical applications. Most existing methods either require a retraining of a given model to achieve robust identification of adversarial attacks or are limited to out-of-distribution sample detection only. In this work, we propose a geometric gradient analysis (GGA) to improve the identification of untrustworthy predictions without retraining of a given model. GGA analyzes the geometry of the loss landscape of neural networks based on the saliency maps of their respective input. We observe considerable differences between the input gradient geometry of trustworthy and untrustworthy predictions. Using these differences, GGA outperforms prior approaches in detecting OOD data and adversarial attacks, including state-of-the-art and adaptive attacks. Code is available at: https://github.com/mad-lab-fau/GGA-Identifying-Untrustworthy-Predictions" @default.
- W3184505055 created "2021-08-02" @default.
- W3184505055 creator A5012198778 @default.
- W3184505055 creator A5023645490 @default.
- W3184505055 creator A5024571000 @default.
- W3184505055 creator A5026483039 @default.
- W3184505055 creator A5028233502 @default.
- W3184505055 creator A5035045595 @default.
- W3184505055 creator A5053758786 @default.
- W3184505055 creator A5069220813 @default.
- W3184505055 date "2021-07-27" @default.
- W3184505055 modified "2023-09-23" @default.
- W3184505055 title "Identifying Untrustworthy Predictions in Neural Networks by Geometric Gradient Analysis" @default.
- W3184505055 hasPublicationYear "2021" @default.
- W3184505055 type Work @default.
- W3184505055 sameAs 3184505055 @default.
- W3184505055 citedByCount "0" @default.
- W3184505055 crossrefType "proceedings-article" @default.
- W3184505055 hasAuthorship W3184505055A5012198778 @default.
- W3184505055 hasAuthorship W3184505055A5023645490 @default.
- W3184505055 hasAuthorship W3184505055A5024571000 @default.
- W3184505055 hasAuthorship W3184505055A5026483039 @default.
- W3184505055 hasAuthorship W3184505055A5028233502 @default.
- W3184505055 hasAuthorship W3184505055A5035045595 @default.
- W3184505055 hasAuthorship W3184505055A5053758786 @default.
- W3184505055 hasAuthorship W3184505055A5069220813 @default.
- W3184505055 hasConcept C116834253 @default.
- W3184505055 hasConcept C119857082 @default.
- W3184505055 hasConcept C144133560 @default.
- W3184505055 hasConcept C153180895 @default.
- W3184505055 hasConcept C153701036 @default.
- W3184505055 hasConcept C154945302 @default.
- W3184505055 hasConcept C155202549 @default.
- W3184505055 hasConcept C177264268 @default.
- W3184505055 hasConcept C185592680 @default.
- W3184505055 hasConcept C198531522 @default.
- W3184505055 hasConcept C199360897 @default.
- W3184505055 hasConcept C2776760102 @default.
- W3184505055 hasConcept C2778712577 @default.
- W3184505055 hasConcept C2984842247 @default.
- W3184505055 hasConcept C37736160 @default.
- W3184505055 hasConcept C38652104 @default.
- W3184505055 hasConcept C41008148 @default.
- W3184505055 hasConcept C43617362 @default.
- W3184505055 hasConcept C50644808 @default.
- W3184505055 hasConcept C59822182 @default.
- W3184505055 hasConcept C86803240 @default.
- W3184505055 hasConceptScore W3184505055C116834253 @default.
- W3184505055 hasConceptScore W3184505055C119857082 @default.
- W3184505055 hasConceptScore W3184505055C144133560 @default.
- W3184505055 hasConceptScore W3184505055C153180895 @default.
- W3184505055 hasConceptScore W3184505055C153701036 @default.
- W3184505055 hasConceptScore W3184505055C154945302 @default.
- W3184505055 hasConceptScore W3184505055C155202549 @default.
- W3184505055 hasConceptScore W3184505055C177264268 @default.
- W3184505055 hasConceptScore W3184505055C185592680 @default.
- W3184505055 hasConceptScore W3184505055C198531522 @default.
- W3184505055 hasConceptScore W3184505055C199360897 @default.
- W3184505055 hasConceptScore W3184505055C2776760102 @default.
- W3184505055 hasConceptScore W3184505055C2778712577 @default.
- W3184505055 hasConceptScore W3184505055C2984842247 @default.
- W3184505055 hasConceptScore W3184505055C37736160 @default.
- W3184505055 hasConceptScore W3184505055C38652104 @default.
- W3184505055 hasConceptScore W3184505055C41008148 @default.
- W3184505055 hasConceptScore W3184505055C43617362 @default.
- W3184505055 hasConceptScore W3184505055C50644808 @default.
- W3184505055 hasConceptScore W3184505055C59822182 @default.
- W3184505055 hasConceptScore W3184505055C86803240 @default.
- W3184505055 hasOpenAccess W3184505055 @default.
- W3184505055 hasRelatedWork W2590523583 @default.
- W3184505055 hasRelatedWork W2891020797 @default.
- W3184505055 hasRelatedWork W2972561187 @default.
- W3184505055 hasRelatedWork W2981077337 @default.
- W3184505055 hasRelatedWork W2991404361 @default.
- W3184505055 hasRelatedWork W2994937424 @default.
- W3184505055 hasRelatedWork W2995319142 @default.
- W3184505055 hasRelatedWork W3082923719 @default.
- W3184505055 hasRelatedWork W3088536850 @default.
- W3184505055 hasRelatedWork W3092696781 @default.
- W3184505055 hasRelatedWork W3093863931 @default.
- W3184505055 hasRelatedWork W3120955712 @default.
- W3184505055 hasRelatedWork W3129912876 @default.
- W3184505055 hasRelatedWork W3130238021 @default.
- W3184505055 hasRelatedWork W3133489578 @default.
- W3184505055 hasRelatedWork W3134761266 @default.
- W3184505055 hasRelatedWork W3151642491 @default.
- W3184505055 hasRelatedWork W3170433571 @default.
- W3184505055 hasRelatedWork W3177448309 @default.
- W3184505055 hasRelatedWork W3183636841 @default.
- W3184505055 isParatext "false" @default.
- W3184505055 isRetracted "false" @default.
- W3184505055 magId "3184505055" @default.
- W3184505055 workType "article" @default.