Matches in SemOpenAlex for { <https://semopenalex.org/work/W4321472312> ?p ?o ?g. }
Showing items 1 to 79 of
79
with 100 items per page.
- W4321472312 abstract "Deep Neural Networks (DNNs) are prone to learn shortcut patterns that damage the generalization of the DNN during deployment. Shortcut Learning is concerning, particularly when the DNNs are applied to safety-critical domains. This paper aims to better understand shortcut learning through the lens of the learning dynamics of the internal neurons during the training process. More specifically, we make the following observations: (1) While previous works treat shortcuts as synonymous with spurious correlations, we emphasize that not all spurious correlations are shortcuts. We show that shortcuts are only those spurious features that are easier than the core features. (2) We build upon this premise and use instance difficulty methods (like Prediction Depth) to quantify easy and to identify this behavior during the training phase. (3) We empirically show that shortcut learning can be detected by observing the learning dynamics of the DNN's early layers, irrespective of the network architecture used. In other words, easy features learned by the initial layers of a DNN early during the training are potential shortcuts. We verify our claims on simulated and real medical imaging data and justify the empirical success of our hypothesis by showing the theoretical connections between Prediction Depth and information-theoretic concepts like V-usable information. Lastly, our experiments show the insufficiency of monitoring only accuracy plots during training (as is common in machine learning pipelines), and we highlight the need for monitoring early training dynamics using example difficulty metrics." @default.
- W4321472312 created "2023-02-22" @default.
- W4321472312 creator A5017974497 @default.
- W4321472312 creator A5022202456 @default.
- W4321472312 creator A5031717623 @default.
- W4321472312 creator A5058743657 @default.
- W4321472312 creator A5073839474 @default.
- W4321472312 date "2023-02-18" @default.
- W4321472312 modified "2023-09-30" @default.
- W4321472312 title "Shortcut Learning Through the Lens of Early Training Dynamics" @default.
- W4321472312 doi "https://doi.org/10.48550/arxiv.2302.09344" @default.
- W4321472312 hasPublicationYear "2023" @default.
- W4321472312 type Work @default.
- W4321472312 citedByCount "0" @default.
- W4321472312 crossrefType "posted-content" @default.
- W4321472312 hasAuthorship W4321472312A5017974497 @default.
- W4321472312 hasAuthorship W4321472312A5022202456 @default.
- W4321472312 hasAuthorship W4321472312A5031717623 @default.
- W4321472312 hasAuthorship W4321472312A5058743657 @default.
- W4321472312 hasAuthorship W4321472312A5073839474 @default.
- W4321472312 hasBestOaLocation W43214723121 @default.
- W4321472312 hasConcept C105339364 @default.
- W4321472312 hasConcept C111919701 @default.
- W4321472312 hasConcept C119857082 @default.
- W4321472312 hasConcept C121332964 @default.
- W4321472312 hasConcept C134306372 @default.
- W4321472312 hasConcept C136764020 @default.
- W4321472312 hasConcept C138885662 @default.
- W4321472312 hasConcept C145912823 @default.
- W4321472312 hasConcept C153294291 @default.
- W4321472312 hasConcept C154945302 @default.
- W4321472312 hasConcept C177148314 @default.
- W4321472312 hasConcept C24890656 @default.
- W4321472312 hasConcept C2777211547 @default.
- W4321472312 hasConcept C2778023277 @default.
- W4321472312 hasConcept C2780615836 @default.
- W4321472312 hasConcept C33923547 @default.
- W4321472312 hasConcept C41008148 @default.
- W4321472312 hasConcept C41895202 @default.
- W4321472312 hasConcept C50644808 @default.
- W4321472312 hasConcept C97256817 @default.
- W4321472312 hasConcept C98045186 @default.
- W4321472312 hasConceptScore W4321472312C105339364 @default.
- W4321472312 hasConceptScore W4321472312C111919701 @default.
- W4321472312 hasConceptScore W4321472312C119857082 @default.
- W4321472312 hasConceptScore W4321472312C121332964 @default.
- W4321472312 hasConceptScore W4321472312C134306372 @default.
- W4321472312 hasConceptScore W4321472312C136764020 @default.
- W4321472312 hasConceptScore W4321472312C138885662 @default.
- W4321472312 hasConceptScore W4321472312C145912823 @default.
- W4321472312 hasConceptScore W4321472312C153294291 @default.
- W4321472312 hasConceptScore W4321472312C154945302 @default.
- W4321472312 hasConceptScore W4321472312C177148314 @default.
- W4321472312 hasConceptScore W4321472312C24890656 @default.
- W4321472312 hasConceptScore W4321472312C2777211547 @default.
- W4321472312 hasConceptScore W4321472312C2778023277 @default.
- W4321472312 hasConceptScore W4321472312C2780615836 @default.
- W4321472312 hasConceptScore W4321472312C33923547 @default.
- W4321472312 hasConceptScore W4321472312C41008148 @default.
- W4321472312 hasConceptScore W4321472312C41895202 @default.
- W4321472312 hasConceptScore W4321472312C50644808 @default.
- W4321472312 hasConceptScore W4321472312C97256817 @default.
- W4321472312 hasConceptScore W4321472312C98045186 @default.
- W4321472312 hasLocation W43214723121 @default.
- W4321472312 hasOpenAccess W4321472312 @default.
- W4321472312 hasPrimaryLocation W43214723121 @default.
- W4321472312 hasRelatedWork W2072735345 @default.
- W4321472312 hasRelatedWork W2952278941 @default.
- W4321472312 hasRelatedWork W2961085424 @default.
- W4321472312 hasRelatedWork W2974706170 @default.
- W4321472312 hasRelatedWork W2989932438 @default.
- W4321472312 hasRelatedWork W3204222276 @default.
- W4321472312 hasRelatedWork W4306674287 @default.
- W4321472312 hasRelatedWork W4312263439 @default.
- W4321472312 hasRelatedWork W1629725936 @default.
- W4321472312 hasRelatedWork W4224009465 @default.
- W4321472312 isParatext "false" @default.
- W4321472312 isRetracted "false" @default.
- W4321472312 workType "article" @default.