Matches in SemOpenAlex for { <https://semopenalex.org/work/W1998656041> ?p ?o ?g. }
Showing items 1 to 78 of
78
with 100 items per page.
- W1998656041 abstract "Recently, sequence level discriminative training methods have been proposed to fine-tune deep neural networks (DNN) after the framelevel cross entropy (CE) training to further improve recognition performance of DNNs. In our previous work, we have proposed a new cluster-based multiple DNNs structure and its parallel training algorithm based on the frame-level cross entropy criterion, which can significantly expedite CE training with multiple GPUs. In this paper, we extend to full sequence training for the multiple DNNs structure for better performance and meanwhile we also consider a partial parallel implementation of sequence training using multiple GPUs for faster training speed. In this work, it is shown that sequence training can be easily extended to multiple DNNs by slightly modifying error signals in output layer. Many implementation steps in sequence training of multiple DNNs can still be parallelized across multiple GPUs for better efficiency. Experiments on the Switchboard task have shown that both frame-level CE training and sequence training of multiple DNNs can lead to massive training speedup with little degradation in recognition performance. Comparing with the state-of-the-art DNN, 4-cluster multiple DNNs model with similar size can achieve more than 7 times faster in CE training and about 1.5 times faster in sequence training when using 4 GPUs." @default.
- W1998656041 created "2016-06-24" @default.
- W1998656041 creator A5008456610 @default.
- W1998656041 creator A5045685496 @default.
- W1998656041 creator A5057227915 @default.
- W1998656041 date "2014-05-01" @default.
- W1998656041 modified "2023-10-16" @default.
- W1998656041 title "Sequence training of multiple deep neural networks for better performance and faster training speed" @default.
- W1998656041 cites W1218987319 @default.
- W1998656041 cites W1498436455 @default.
- W1998656041 cites W1987238397 @default.
- W1998656041 cites W2012257340 @default.
- W1998656041 cites W2020144989 @default.
- W1998656041 cites W2033565080 @default.
- W1998656041 cites W2058641082 @default.
- W1998656041 cites W2063689849 @default.
- W1998656041 cites W2081884280 @default.
- W1998656041 cites W2087402357 @default.
- W1998656041 cites W2112984492 @default.
- W1998656041 cites W2403195671 @default.
- W1998656041 cites W38527073 @default.
- W1998656041 doi "https://doi.org/10.1109/icassp.2014.6854680" @default.
- W1998656041 hasPublicationYear "2014" @default.
- W1998656041 type Work @default.
- W1998656041 sameAs 1998656041 @default.
- W1998656041 citedByCount "15" @default.
- W1998656041 countsByYear W19986560412014 @default.
- W1998656041 countsByYear W19986560412015 @default.
- W1998656041 countsByYear W19986560412016 @default.
- W1998656041 countsByYear W19986560412017 @default.
- W1998656041 countsByYear W19986560412019 @default.
- W1998656041 countsByYear W19986560412020 @default.
- W1998656041 countsByYear W19986560412022 @default.
- W1998656041 crossrefType "proceedings-article" @default.
- W1998656041 hasAuthorship W1998656041A5008456610 @default.
- W1998656041 hasAuthorship W1998656041A5045685496 @default.
- W1998656041 hasAuthorship W1998656041A5057227915 @default.
- W1998656041 hasConcept C119857082 @default.
- W1998656041 hasConcept C121332964 @default.
- W1998656041 hasConcept C153294291 @default.
- W1998656041 hasConcept C154945302 @default.
- W1998656041 hasConcept C155032097 @default.
- W1998656041 hasConcept C2777211547 @default.
- W1998656041 hasConcept C2778112365 @default.
- W1998656041 hasConcept C41008148 @default.
- W1998656041 hasConcept C50644808 @default.
- W1998656041 hasConcept C51632099 @default.
- W1998656041 hasConcept C54355233 @default.
- W1998656041 hasConcept C86803240 @default.
- W1998656041 hasConceptScore W1998656041C119857082 @default.
- W1998656041 hasConceptScore W1998656041C121332964 @default.
- W1998656041 hasConceptScore W1998656041C153294291 @default.
- W1998656041 hasConceptScore W1998656041C154945302 @default.
- W1998656041 hasConceptScore W1998656041C155032097 @default.
- W1998656041 hasConceptScore W1998656041C2777211547 @default.
- W1998656041 hasConceptScore W1998656041C2778112365 @default.
- W1998656041 hasConceptScore W1998656041C41008148 @default.
- W1998656041 hasConceptScore W1998656041C50644808 @default.
- W1998656041 hasConceptScore W1998656041C51632099 @default.
- W1998656041 hasConceptScore W1998656041C54355233 @default.
- W1998656041 hasConceptScore W1998656041C86803240 @default.
- W1998656041 hasLocation W19986560411 @default.
- W1998656041 hasOpenAccess W1998656041 @default.
- W1998656041 hasPrimaryLocation W19986560411 @default.
- W1998656041 hasRelatedWork W1587050645 @default.
- W1998656041 hasRelatedWork W2056364562 @default.
- W1998656041 hasRelatedWork W2062010226 @default.
- W1998656041 hasRelatedWork W2101017737 @default.
- W1998656041 hasRelatedWork W2106973078 @default.
- W1998656041 hasRelatedWork W2159443810 @default.
- W1998656041 hasRelatedWork W303964064 @default.
- W1998656041 hasRelatedWork W4234997060 @default.
- W1998656041 hasRelatedWork W4238261984 @default.
- W1998656041 hasRelatedWork W1629725936 @default.
- W1998656041 isParatext "false" @default.
- W1998656041 isRetracted "false" @default.
- W1998656041 magId "1998656041" @default.
- W1998656041 workType "article" @default.