FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2973118234> ?p ?o ?g. }

• W2973118234 endingPage "62447" @default.
• W2973118234 startingPage "62439" @default.
• W2973118234 abstract "Training of large scale deep neural networks with distributed implementations is an effective way to improve the efficiency. However, high network communication cost for synchronizing gradients and parameters is a major bottleneck in distributed training. In this work, we propose an asynchronous training algorithm based on sampling and mean fusion for distributed recurrent neural network (RNN). In distributed RNN, multiple distributed neuron nodes and an interaction node work together to implement the training. The synchronization overhead is reduced by a unique asynchronous sampling strategy amongst the distributed neuron nodes. Then, in order to make up for the accuracy loss caused by the asynchronous parameter update, a mean fusion algorithm is proposed, where the interaction node averages all local parameters from the distributed neurons. We mathematically prove the convergence of the proposed algorithm. Experimental verification is performed on two language modeling benchmark datasets. The results demonstrate significant speed gains for distributed RNN, while the accuracy loss is less than 1% on average." @default.
• W2973118234 created "2019-09-19" @default.
• W2973118234 creator A5003123115 @default.
• W2973118234 creator A5003761302 @default.
• W2973118234 creator A5012385922 @default.
• W2973118234 creator A5028022474 @default.
• W2973118234 date "2020-01-01" @default.
• W2973118234 modified "2023-09-27" @default.
• W2973118234 title "The Asynchronous Training Algorithm Based on Sampling and Mean Fusion for Distributed RNN" @default.
• W2973118234 cites W1603614877 @default.
• W2973118234 cites W1632114991 @default.
• W2973118234 cites W1895577753 @default.
• W2973118234 cites W1947481528 @default.
• W2973118234 cites W2086161653 @default.
• W2973118234 cites W2097117768 @default.
• W2973118234 cites W2115613106 @default.
• W2973118234 cites W2157331557 @default.
• W2973118234 cites W2160815625 @default.
• W2973118234 cites W2731809466 @default.
• W2973118234 doi "https://doi.org/10.1109/access.2019.2939851" @default.
• W2973118234 hasPublicationYear "2020" @default.
• W2973118234 type Work @default.
• W2973118234 sameAs 2973118234 @default.
• W2973118234 citedByCount "2" @default.
• W2973118234 countsByYear W29731182342020 @default.
• W2973118234 countsByYear W29731182342022 @default.
• W2973118234 crossrefType "journal-article" @default.
• W2973118234 hasAuthorship W2973118234A5003123115 @default.
• W2973118234 hasAuthorship W2973118234A5003761302 @default.
• W2973118234 hasAuthorship W2973118234A5012385922 @default.
• W2973118234 hasAuthorship W2973118234A5028022474 @default.
• W2973118234 hasBestOaLocation W29731182341 @default.
• W2973118234 hasConcept C111919701 @default.
• W2973118234 hasConcept C11413529 @default.
• W2973118234 hasConcept C120314980 @default.
• W2973118234 hasConcept C127162648 @default.
• W2973118234 hasConcept C127413603 @default.
• W2973118234 hasConcept C130120984 @default.
• W2973118234 hasConcept C13280743 @default.
• W2973118234 hasConcept C147168706 @default.
• W2973118234 hasConcept C149635348 @default.
• W2973118234 hasConcept C151319957 @default.
• W2973118234 hasConcept C154945302 @default.
• W2973118234 hasConcept C162324750 @default.
• W2973118234 hasConcept C185798385 @default.
• W2973118234 hasConcept C205649164 @default.
• W2973118234 hasConcept C2777303404 @default.
• W2973118234 hasConcept C2778562939 @default.
• W2973118234 hasConcept C2779960059 @default.
• W2973118234 hasConcept C2780513914 @default.
• W2973118234 hasConcept C31258907 @default.
• W2973118234 hasConcept C41008148 @default.
• W2973118234 hasConcept C50522688 @default.
• W2973118234 hasConcept C50644808 @default.
• W2973118234 hasConcept C62611344 @default.
• W2973118234 hasConcept C66938386 @default.
• W2973118234 hasConceptScore W2973118234C111919701 @default.
• W2973118234 hasConceptScore W2973118234C11413529 @default.
• W2973118234 hasConceptScore W2973118234C120314980 @default.
• W2973118234 hasConceptScore W2973118234C127162648 @default.
• W2973118234 hasConceptScore W2973118234C127413603 @default.
• W2973118234 hasConceptScore W2973118234C130120984 @default.
• W2973118234 hasConceptScore W2973118234C13280743 @default.
• W2973118234 hasConceptScore W2973118234C147168706 @default.
• W2973118234 hasConceptScore W2973118234C149635348 @default.
• W2973118234 hasConceptScore W2973118234C151319957 @default.
• W2973118234 hasConceptScore W2973118234C154945302 @default.
• W2973118234 hasConceptScore W2973118234C162324750 @default.
• W2973118234 hasConceptScore W2973118234C185798385 @default.
• W2973118234 hasConceptScore W2973118234C205649164 @default.
• W2973118234 hasConceptScore W2973118234C2777303404 @default.
• W2973118234 hasConceptScore W2973118234C2778562939 @default.
• W2973118234 hasConceptScore W2973118234C2779960059 @default.
• W2973118234 hasConceptScore W2973118234C2780513914 @default.
• W2973118234 hasConceptScore W2973118234C31258907 @default.
• W2973118234 hasConceptScore W2973118234C41008148 @default.
• W2973118234 hasConceptScore W2973118234C50522688 @default.
• W2973118234 hasConceptScore W2973118234C50644808 @default.
• W2973118234 hasConceptScore W2973118234C62611344 @default.
• W2973118234 hasConceptScore W2973118234C66938386 @default.
• W2973118234 hasFunder F4320321001 @default.
• W2973118234 hasFunder F4320321543 @default.
• W2973118234 hasLocation W29731182341 @default.
• W2973118234 hasOpenAccess W2973118234 @default.
• W2973118234 hasPrimaryLocation W29731182341 @default.
• W2973118234 hasRelatedWork W1485627940 @default.
• W2973118234 hasRelatedWork W1966600221 @default.
• W2973118234 hasRelatedWork W2071774063 @default.
• W2973118234 hasRelatedWork W2131630752 @default.
• W2973118234 hasRelatedWork W2133123956 @default.
• W2973118234 hasRelatedWork W2147711412 @default.
• W2973118234 hasRelatedWork W2971772242 @default.
• W2973118234 hasRelatedWork W3128207141 @default.
• W2973118234 hasRelatedWork W94000989 @default.
• W2973118234 hasRelatedWork W2516240846 @default.
• W2973118234 hasVolume "8" @default.
• W2973118234 isParatext "false" @default.
• W2973118234 isRetracted "false" @default.

• next