Matches in SemOpenAlex for { <https://semopenalex.org/work/W2981884310> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W2981884310 abstract "Multi-exit architectures, in which a stack of processing layers is interleaved with early output layers, allow the processing of a test example to stop early and thus save computation time and/or energy. In this work, we propose a new training procedure for multi-exit architectures based on the principle of knowledge distillation. The method encourages early exits to mimic later, more accurate exits, by matching their probability outputs. Experiments on CIFAR100 and ImageNet show that distillation-based training significantly improves the accuracy of early exits while maintaining state-of-the-art accuracy for late ones. The method is particularly beneficial when training data is limited and also allows a straight-forward extension to semi-supervised learning, i.e. make use also of unlabeled data at training time. Moreover, it takes only a few lines to implement and imposes almost no computational overhead at training time, and none at all at test time." @default.
- W2981884310 created "2019-11-01" @default.
- W2981884310 creator A5034739012 @default.
- W2981884310 creator A5068085751 @default.
- W2981884310 date "2019-10-01" @default.
- W2981884310 modified "2023-10-05" @default.
- W2981884310 title "Distillation-Based Training for Multi-Exit Architectures" @default.
- W2981884310 cites W1602122992 @default.
- W2981884310 cites W2048500237 @default.
- W2981884310 cites W2108598243 @default.
- W2981884310 cites W2109483014 @default.
- W2981884310 cites W2152497905 @default.
- W2981884310 cites W2164598857 @default.
- W2981884310 cites W2194775991 @default.
- W2981884310 cites W2294370754 @default.
- W2981884310 cites W2592335154 @default.
- W2981884310 cites W2962677625 @default.
- W2981884310 cites W2962737704 @default.
- W2981884310 cites W2962944050 @default.
- W2981884310 cites W2963687836 @default.
- W2981884310 cites W2963785012 @default.
- W2981884310 cites W2964082701 @default.
- W2981884310 cites W2964189064 @default.
- W2981884310 cites W3097096317 @default.
- W2981884310 cites W38811222 @default.
- W2981884310 cites W845365781 @default.
- W2981884310 doi "https://doi.org/10.1109/iccv.2019.00144" @default.
- W2981884310 hasPublicationYear "2019" @default.
- W2981884310 type Work @default.
- W2981884310 sameAs 2981884310 @default.
- W2981884310 citedByCount "79" @default.
- W2981884310 countsByYear W29818843102019 @default.
- W2981884310 countsByYear W29818843102020 @default.
- W2981884310 countsByYear W29818843102021 @default.
- W2981884310 countsByYear W29818843102022 @default.
- W2981884310 countsByYear W29818843102023 @default.
- W2981884310 crossrefType "proceedings-article" @default.
- W2981884310 hasAuthorship W2981884310A5034739012 @default.
- W2981884310 hasAuthorship W2981884310A5068085751 @default.
- W2981884310 hasConcept C105795698 @default.
- W2981884310 hasConcept C111919701 @default.
- W2981884310 hasConcept C11413529 @default.
- W2981884310 hasConcept C119857082 @default.
- W2981884310 hasConcept C121332964 @default.
- W2981884310 hasConcept C153294291 @default.
- W2981884310 hasConcept C154945302 @default.
- W2981884310 hasConcept C165064840 @default.
- W2981884310 hasConcept C16910744 @default.
- W2981884310 hasConcept C178790620 @default.
- W2981884310 hasConcept C185592680 @default.
- W2981884310 hasConcept C199360897 @default.
- W2981884310 hasConcept C204030448 @default.
- W2981884310 hasConcept C2777211547 @default.
- W2981884310 hasConcept C2779960059 @default.
- W2981884310 hasConcept C33923547 @default.
- W2981884310 hasConcept C41008148 @default.
- W2981884310 hasConcept C45374587 @default.
- W2981884310 hasConcept C9395851 @default.
- W2981884310 hasConceptScore W2981884310C105795698 @default.
- W2981884310 hasConceptScore W2981884310C111919701 @default.
- W2981884310 hasConceptScore W2981884310C11413529 @default.
- W2981884310 hasConceptScore W2981884310C119857082 @default.
- W2981884310 hasConceptScore W2981884310C121332964 @default.
- W2981884310 hasConceptScore W2981884310C153294291 @default.
- W2981884310 hasConceptScore W2981884310C154945302 @default.
- W2981884310 hasConceptScore W2981884310C165064840 @default.
- W2981884310 hasConceptScore W2981884310C16910744 @default.
- W2981884310 hasConceptScore W2981884310C178790620 @default.
- W2981884310 hasConceptScore W2981884310C185592680 @default.
- W2981884310 hasConceptScore W2981884310C199360897 @default.
- W2981884310 hasConceptScore W2981884310C204030448 @default.
- W2981884310 hasConceptScore W2981884310C2777211547 @default.
- W2981884310 hasConceptScore W2981884310C2779960059 @default.
- W2981884310 hasConceptScore W2981884310C33923547 @default.
- W2981884310 hasConceptScore W2981884310C41008148 @default.
- W2981884310 hasConceptScore W2981884310C45374587 @default.
- W2981884310 hasConceptScore W2981884310C9395851 @default.
- W2981884310 hasLocation W29818843101 @default.
- W2981884310 hasOpenAccess W2981884310 @default.
- W2981884310 hasPrimaryLocation W29818843101 @default.
- W2981884310 hasRelatedWork W2028024605 @default.
- W2981884310 hasRelatedWork W2361879592 @default.
- W2981884310 hasRelatedWork W2961085424 @default.
- W2981884310 hasRelatedWork W3046775127 @default.
- W2981884310 hasRelatedWork W3170094116 @default.
- W2981884310 hasRelatedWork W4285260836 @default.
- W2981884310 hasRelatedWork W4286629047 @default.
- W2981884310 hasRelatedWork W4306321456 @default.
- W2981884310 hasRelatedWork W4306674287 @default.
- W2981884310 hasRelatedWork W4224009465 @default.
- W2981884310 isParatext "false" @default.
- W2981884310 isRetracted "false" @default.
- W2981884310 magId "2981884310" @default.
- W2981884310 workType "article" @default.