Matches in SemOpenAlex for { <https://semopenalex.org/work/W2798722023> ?p ?o ?g. }
Showing items 1 to 88 of
88
with 100 items per page.
- W2798722023 abstract "There is growing interest in improving the design of deep network architectures to be both accurate and low cost. This paper explores semantic specialization as a mechanism for improving the computational efficiency (accuracy-per-unit-cost) of inference in the context of image classification. Specifically, we propose a network architecture template called HydraNet, which enables state-of-the-art architectures for image classification to be transformed into dynamic architectures which exploit conditional execution for efficient inference. HydraNets are wide networks containing distinct components specialized to compute features for visually similar classes, but they retain efficiency by dynamically selecting only a small number of components to evaluate for any one input image. This design is made possible by a soft gating mechanism that encourages component specialization during training and accurately performs component selection during inference. We evaluate the HydraNet approach on both the CIFAR-100 and ImageNet classification tasks. On CIFAR, applying the HydraNet template to the ResNet and DenseNet family of models reduces inference cost by 2-4× while retaining the accuracy of the baseline architectures. On ImageNet, applying the HydraNet template improves accuracy up to 2.5% when compared to an efficient baseline architecture with similar inference cost." @default.
- W2798722023 created "2018-05-07" @default.
- W2798722023 creator A5006412542 @default.
- W2798722023 creator A5021878400 @default.
- W2798722023 creator A5037444018 @default.
- W2798722023 creator A5088661550 @default.
- W2798722023 date "2018-06-01" @default.
- W2798722023 modified "2023-10-16" @default.
- W2798722023 title "HydraNets: Specialized Dynamic Architectures for Efficient Inference" @default.
- W2798722023 cites W1978962787 @default.
- W2798722023 cites W2097117768 @default.
- W2798722023 cites W2108598243 @default.
- W2798722023 cites W2117539524 @default.
- W2798722023 cites W2164598857 @default.
- W2798722023 cites W2194775991 @default.
- W2798722023 cites W2220384803 @default.
- W2798722023 cites W2285660444 @default.
- W2798722023 cites W2606722458 @default.
- W2798722023 cites W2962716332 @default.
- W2798722023 cites W2963446712 @default.
- W2798722023 cites W2964118342 @default.
- W2798722023 doi "https://doi.org/10.1109/cvpr.2018.00843" @default.
- W2798722023 hasPublicationYear "2018" @default.
- W2798722023 type Work @default.
- W2798722023 sameAs 2798722023 @default.
- W2798722023 citedByCount "50" @default.
- W2798722023 countsByYear W27987220232018 @default.
- W2798722023 countsByYear W27987220232019 @default.
- W2798722023 countsByYear W27987220232020 @default.
- W2798722023 countsByYear W27987220232021 @default.
- W2798722023 countsByYear W27987220232022 @default.
- W2798722023 countsByYear W27987220232023 @default.
- W2798722023 crossrefType "proceedings-article" @default.
- W2798722023 hasAuthorship W2798722023A5006412542 @default.
- W2798722023 hasAuthorship W2798722023A5021878400 @default.
- W2798722023 hasAuthorship W2798722023A5037444018 @default.
- W2798722023 hasAuthorship W2798722023A5088661550 @default.
- W2798722023 hasConcept C119857082 @default.
- W2798722023 hasConcept C121332964 @default.
- W2798722023 hasConcept C123657996 @default.
- W2798722023 hasConcept C142362112 @default.
- W2798722023 hasConcept C151730666 @default.
- W2798722023 hasConcept C153180895 @default.
- W2798722023 hasConcept C153349607 @default.
- W2798722023 hasConcept C154945302 @default.
- W2798722023 hasConcept C165696696 @default.
- W2798722023 hasConcept C168167062 @default.
- W2798722023 hasConcept C193415008 @default.
- W2798722023 hasConcept C2776214188 @default.
- W2798722023 hasConcept C2779343474 @default.
- W2798722023 hasConcept C38652104 @default.
- W2798722023 hasConcept C41008148 @default.
- W2798722023 hasConcept C86803240 @default.
- W2798722023 hasConcept C97355855 @default.
- W2798722023 hasConceptScore W2798722023C119857082 @default.
- W2798722023 hasConceptScore W2798722023C121332964 @default.
- W2798722023 hasConceptScore W2798722023C123657996 @default.
- W2798722023 hasConceptScore W2798722023C142362112 @default.
- W2798722023 hasConceptScore W2798722023C151730666 @default.
- W2798722023 hasConceptScore W2798722023C153180895 @default.
- W2798722023 hasConceptScore W2798722023C153349607 @default.
- W2798722023 hasConceptScore W2798722023C154945302 @default.
- W2798722023 hasConceptScore W2798722023C165696696 @default.
- W2798722023 hasConceptScore W2798722023C168167062 @default.
- W2798722023 hasConceptScore W2798722023C193415008 @default.
- W2798722023 hasConceptScore W2798722023C2776214188 @default.
- W2798722023 hasConceptScore W2798722023C2779343474 @default.
- W2798722023 hasConceptScore W2798722023C38652104 @default.
- W2798722023 hasConceptScore W2798722023C41008148 @default.
- W2798722023 hasConceptScore W2798722023C86803240 @default.
- W2798722023 hasConceptScore W2798722023C97355855 @default.
- W2798722023 hasLocation W27987220231 @default.
- W2798722023 hasOpenAccess W2798722023 @default.
- W2798722023 hasPrimaryLocation W27987220231 @default.
- W2798722023 hasRelatedWork W2364531466 @default.
- W2798722023 hasRelatedWork W2382623646 @default.
- W2798722023 hasRelatedWork W2898291644 @default.
- W2798722023 hasRelatedWork W2963058055 @default.
- W2798722023 hasRelatedWork W2963456518 @default.
- W2798722023 hasRelatedWork W3011666637 @default.
- W2798722023 hasRelatedWork W3104224589 @default.
- W2798722023 hasRelatedWork W396164270 @default.
- W2798722023 hasRelatedWork W4206079793 @default.
- W2798722023 hasRelatedWork W4297845793 @default.
- W2798722023 isParatext "false" @default.
- W2798722023 isRetracted "false" @default.
- W2798722023 magId "2798722023" @default.
- W2798722023 workType "article" @default.