Matches in SemOpenAlex for { <https://semopenalex.org/work/W3176808018> ?p ?o ?g. }
- W3176808018 abstract "We propose multirate training of neural networks: partitioning neural network parameters into fast and slow parts which are trained simultaneously using different learning rates. By choosing appropriate partitionings we can obtain large computational speed-ups for transfer learning tasks. We show that for various transfer learning applications in vision and NLP we can fine-tune deep neural networks in almost half the time, without reducing the generalization performance of the resulting model. We also discuss other splitting choices for the neural network parameters which are beneficial in enhancing generalization performance in settings where neural networks are trained from scratch. Finally, we propose an additional multirate technique which can learn different features present in the data by training the full network on different time scales simultaneously. The benefits of using this approach are illustrated for ResNet architectures on image data. Our paper unlocks the potential of using multirate techniques for neural network training and provides many starting points for future work in this area." @default.
- W3176808018 created "2021-07-05" @default.
- W3176808018 creator A5002315948 @default.
- W3176808018 creator A5068323517 @default.
- W3176808018 date "2021-06-20" @default.
- W3176808018 modified "2023-09-27" @default.
- W3176808018 title "Multirate Training of Neural Networks." @default.
- W3176808018 cites W1488765129 @default.
- W3176808018 cites W1525859397 @default.
- W3176808018 cites W1836465849 @default.
- W3176808018 cites W1966173660 @default.
- W3176808018 cites W2017835622 @default.
- W3176808018 cites W2018415001 @default.
- W3176808018 cites W2035851348 @default.
- W3176808018 cites W2087654468 @default.
- W3176808018 cites W2095705004 @default.
- W3176808018 cites W2104049095 @default.
- W3176808018 cites W2136804894 @default.
- W3176808018 cites W2167433878 @default.
- W3176808018 cites W2170973209 @default.
- W3176808018 cites W2194775991 @default.
- W3176808018 cites W2251939518 @default.
- W3176808018 cites W2510153535 @default.
- W3176808018 cites W2518177752 @default.
- W3176808018 cites W2535697732 @default.
- W3176808018 cites W2752512710 @default.
- W3176808018 cites W2757910899 @default.
- W3176808018 cites W2798812533 @default.
- W3176808018 cites W2891172060 @default.
- W3176808018 cites W2899771611 @default.
- W3176808018 cites W2913190747 @default.
- W3176808018 cites W2945313542 @default.
- W3176808018 cites W2949667497 @default.
- W3176808018 cites W2951584298 @default.
- W3176808018 cites W2962734576 @default.
- W3176808018 cites W2963173418 @default.
- W3176808018 cites W2963341956 @default.
- W3176808018 cites W2967659330 @default.
- W3176808018 cites W2970388773 @default.
- W3176808018 cites W2970854840 @default.
- W3176808018 cites W2991391304 @default.
- W3176808018 cites W2995435108 @default.
- W3176808018 cites W2996056314 @default.
- W3176808018 cites W3034669169 @default.
- W3176808018 cites W3098033339 @default.
- W3176808018 cites W3118608800 @default.
- W3176808018 cites W4919037 @default.
- W3176808018 hasPublicationYear "2021" @default.
- W3176808018 type Work @default.
- W3176808018 sameAs 3176808018 @default.
- W3176808018 citedByCount "0" @default.
- W3176808018 crossrefType "posted-content" @default.
- W3176808018 hasAuthorship W3176808018A5002315948 @default.
- W3176808018 hasAuthorship W3176808018A5068323517 @default.
- W3176808018 hasConcept C108583219 @default.
- W3176808018 hasConcept C111919701 @default.
- W3176808018 hasConcept C119857082 @default.
- W3176808018 hasConcept C121332964 @default.
- W3176808018 hasConcept C134306372 @default.
- W3176808018 hasConcept C150899416 @default.
- W3176808018 hasConcept C153294291 @default.
- W3176808018 hasConcept C154945302 @default.
- W3176808018 hasConcept C175202392 @default.
- W3176808018 hasConcept C177148314 @default.
- W3176808018 hasConcept C2777211547 @default.
- W3176808018 hasConcept C2781235140 @default.
- W3176808018 hasConcept C2984842247 @default.
- W3176808018 hasConcept C33923547 @default.
- W3176808018 hasConcept C41008148 @default.
- W3176808018 hasConcept C50644808 @default.
- W3176808018 hasConcept C51632099 @default.
- W3176808018 hasConceptScore W3176808018C108583219 @default.
- W3176808018 hasConceptScore W3176808018C111919701 @default.
- W3176808018 hasConceptScore W3176808018C119857082 @default.
- W3176808018 hasConceptScore W3176808018C121332964 @default.
- W3176808018 hasConceptScore W3176808018C134306372 @default.
- W3176808018 hasConceptScore W3176808018C150899416 @default.
- W3176808018 hasConceptScore W3176808018C153294291 @default.
- W3176808018 hasConceptScore W3176808018C154945302 @default.
- W3176808018 hasConceptScore W3176808018C175202392 @default.
- W3176808018 hasConceptScore W3176808018C177148314 @default.
- W3176808018 hasConceptScore W3176808018C2777211547 @default.
- W3176808018 hasConceptScore W3176808018C2781235140 @default.
- W3176808018 hasConceptScore W3176808018C2984842247 @default.
- W3176808018 hasConceptScore W3176808018C33923547 @default.
- W3176808018 hasConceptScore W3176808018C41008148 @default.
- W3176808018 hasConceptScore W3176808018C50644808 @default.
- W3176808018 hasConceptScore W3176808018C51632099 @default.
- W3176808018 hasLocation W31768080181 @default.
- W3176808018 hasOpenAccess W3176808018 @default.
- W3176808018 hasPrimaryLocation W31768080181 @default.
- W3176808018 hasRelatedWork W1864296793 @default.
- W3176808018 hasRelatedWork W2077700203 @default.
- W3176808018 hasRelatedWork W2143859313 @default.
- W3176808018 hasRelatedWork W2365826670 @default.
- W3176808018 hasRelatedWork W2468983453 @default.
- W3176808018 hasRelatedWork W2493224314 @default.
- W3176808018 hasRelatedWork W2520127694 @default.
- W3176808018 hasRelatedWork W2525954470 @default.
- W3176808018 hasRelatedWork W2800011138 @default.