Matches in SemOpenAlex for { <https://semopenalex.org/work/W4285164920> ?p ?o ?g. }
Showing items 1 to 88 of
88
with 100 items per page.
- W4285164920 endingPage "357" @default.
- W4285164920 startingPage "345" @default.
- W4285164920 abstract "AbstractNeural networks are more expressive when they have multiple layers. In turn, conventional training methods are only successful if the depth does not lead to numerical issues such as exploding or vanishing gradients, which occur less frequently when the layers are sufficiently wide. However, increasing width to attain greater depth entails the use of heavier computational resources and leads to overparameterized models. These subsequent issues have been partially addressed by model compression methods such as quantization and pruning, some of which relying on normalization-based regularization of the loss function to make the effect of most parameters negligible. In this work, we propose instead to use regularization for preventing neurons from dying or becoming linear, a technique which we denote as jumpstart regularization. In comparison to conventional training, we obtain neural networks that are thinner, deeper, and—most importantly—more parameter-efficient.KeywordsDeep learningModel compressionReLU networks" @default.
- W4285164920 created "2022-07-14" @default.
- W4285164920 creator A5017504613 @default.
- W4285164920 creator A5031445379 @default.
- W4285164920 creator A5038075462 @default.
- W4285164920 creator A5090793477 @default.
- W4285164920 creator A5091099873 @default.
- W4285164920 date "2022-01-01" @default.
- W4285164920 modified "2023-09-27" @default.
- W4285164920 title "Training Thinner and Deeper Neural Networks: Jumpstart Regularization" @default.
- W4285164920 cites W1498436455 @default.
- W4285164920 cites W1572063013 @default.
- W4285164920 cites W1677182931 @default.
- W4285164920 cites W183625566 @default.
- W4285164920 cites W1899504021 @default.
- W4285164920 cites W197865394 @default.
- W4285164920 cites W1994616650 @default.
- W4285164920 cites W20283819 @default.
- W4285164920 cites W2064675550 @default.
- W4285164920 cites W2103496339 @default.
- W4285164920 cites W2107878631 @default.
- W4285164920 cites W2112796928 @default.
- W4285164920 cites W2919115771 @default.
- W4285164920 cites W2943191253 @default.
- W4285164920 cites W2963446712 @default.
- W4285164920 cites W2963809228 @default.
- W4285164920 cites W2964137095 @default.
- W4285164920 cites W2973514284 @default.
- W4285164920 cites W2998329861 @default.
- W4285164920 cites W3089719735 @default.
- W4285164920 cites W3133702157 @default.
- W4285164920 cites W4236362309 @default.
- W4285164920 doi "https://doi.org/10.1007/978-3-031-08011-1_23" @default.
- W4285164920 hasPublicationYear "2022" @default.
- W4285164920 type Work @default.
- W4285164920 citedByCount "0" @default.
- W4285164920 crossrefType "book-chapter" @default.
- W4285164920 hasAuthorship W4285164920A5017504613 @default.
- W4285164920 hasAuthorship W4285164920A5031445379 @default.
- W4285164920 hasAuthorship W4285164920A5038075462 @default.
- W4285164920 hasAuthorship W4285164920A5090793477 @default.
- W4285164920 hasAuthorship W4285164920A5091099873 @default.
- W4285164920 hasBestOaLocation W42851649202 @default.
- W4285164920 hasConcept C11413529 @default.
- W4285164920 hasConcept C119857082 @default.
- W4285164920 hasConcept C126255220 @default.
- W4285164920 hasConcept C136886441 @default.
- W4285164920 hasConcept C144024400 @default.
- W4285164920 hasConcept C154945302 @default.
- W4285164920 hasConcept C19165224 @default.
- W4285164920 hasConcept C2776135515 @default.
- W4285164920 hasConcept C28855332 @default.
- W4285164920 hasConcept C2984842247 @default.
- W4285164920 hasConcept C33923547 @default.
- W4285164920 hasConcept C41008148 @default.
- W4285164920 hasConcept C50644808 @default.
- W4285164920 hasConceptScore W4285164920C11413529 @default.
- W4285164920 hasConceptScore W4285164920C119857082 @default.
- W4285164920 hasConceptScore W4285164920C126255220 @default.
- W4285164920 hasConceptScore W4285164920C136886441 @default.
- W4285164920 hasConceptScore W4285164920C144024400 @default.
- W4285164920 hasConceptScore W4285164920C154945302 @default.
- W4285164920 hasConceptScore W4285164920C19165224 @default.
- W4285164920 hasConceptScore W4285164920C2776135515 @default.
- W4285164920 hasConceptScore W4285164920C28855332 @default.
- W4285164920 hasConceptScore W4285164920C2984842247 @default.
- W4285164920 hasConceptScore W4285164920C33923547 @default.
- W4285164920 hasConceptScore W4285164920C41008148 @default.
- W4285164920 hasConceptScore W4285164920C50644808 @default.
- W4285164920 hasLocation W42851649201 @default.
- W4285164920 hasLocation W42851649202 @default.
- W4285164920 hasOpenAccess W4285164920 @default.
- W4285164920 hasPrimaryLocation W42851649201 @default.
- W4285164920 hasRelatedWork W2516800609 @default.
- W4285164920 hasRelatedWork W2594653239 @default.
- W4285164920 hasRelatedWork W2891335318 @default.
- W4285164920 hasRelatedWork W2961262370 @default.
- W4285164920 hasRelatedWork W2974456200 @default.
- W4285164920 hasRelatedWork W3082154786 @default.
- W4285164920 hasRelatedWork W4288853838 @default.
- W4285164920 hasRelatedWork W4297931150 @default.
- W4285164920 hasRelatedWork W4301965327 @default.
- W4285164920 hasRelatedWork W1629725936 @default.
- W4285164920 isParatext "false" @default.
- W4285164920 isRetracted "false" @default.
- W4285164920 workType "book-chapter" @default.