Matches in SemOpenAlex for { <https://semopenalex.org/work/W3204348672> ?p ?o ?g. }
Showing items 1 to 87 of
87
with 100 items per page.
- W3204348672 abstract "Large-scale distributed training has recently been proposed as a solution to speed-up the training of deep neural networks on huge datasets. Distributed training, however, entails high communication rates for gradient exchange among computing nodes and requires expensive high-bandwidth network infrastructure. Various gradient compression methods have been proposed to overcome this limitation, including sparsification, quantization, and entropy encoding of the gradients. However, most existing methods leverage only the intra-node information redundancy, that is, they compress gradients at each node independently. In contrast, we advocate that the gradients across the nodes are correlated and propose a method to leverage this inter-node redundancy to obtain higher compression rates. In this work, we propose the Learned Gradient Compression (LGC) framework to reduce communication rates within a distributed training with the parameter server communication protocol. Our framework leverages an autoencoder to capture the common information in the gradients of the distributed nodes and eliminate the transmission of redundant information. Our experiments show that the proposed approach achieves significantly higher gradient compression ratios than state-of-the-art approaches like DGC and ScaleCom." @default.
- W3204348672 created "2021-10-11" @default.
- W3204348672 creator A5043377240 @default.
- W3204348672 creator A5043511500 @default.
- W3204348672 creator A5085612711 @default.
- W3204348672 creator A5091029466 @default.
- W3204348672 date "2021-08-23" @default.
- W3204348672 modified "2023-09-27" @default.
- W3204348672 title "Autoencoder-Based Gradient Compression for Distributed Training" @default.
- W3204348672 cites W1533527630 @default.
- W3204348672 cites W1686810756 @default.
- W3204348672 cites W2108598243 @default.
- W3204348672 cites W2168231600 @default.
- W3204348672 cites W2194775991 @default.
- W3204348672 cites W2405578611 @default.
- W3204348672 cites W2560023338 @default.
- W3204348672 cites W2963803379 @default.
- W3204348672 cites W2970895312 @default.
- W3204348672 cites W2971064744 @default.
- W3204348672 cites W3037585619 @default.
- W3204348672 doi "https://doi.org/10.23919/eusipco54536.2021.9616078" @default.
- W3204348672 hasPublicationYear "2021" @default.
- W3204348672 type Work @default.
- W3204348672 sameAs 3204348672 @default.
- W3204348672 citedByCount "0" @default.
- W3204348672 crossrefType "proceedings-article" @default.
- W3204348672 hasAuthorship W3204348672A5043377240 @default.
- W3204348672 hasAuthorship W3204348672A5043511500 @default.
- W3204348672 hasAuthorship W3204348672A5085612711 @default.
- W3204348672 hasAuthorship W3204348672A5091029466 @default.
- W3204348672 hasConcept C101738243 @default.
- W3204348672 hasConcept C106301342 @default.
- W3204348672 hasConcept C111919701 @default.
- W3204348672 hasConcept C113775141 @default.
- W3204348672 hasConcept C11413529 @default.
- W3204348672 hasConcept C118505674 @default.
- W3204348672 hasConcept C120314980 @default.
- W3204348672 hasConcept C121332964 @default.
- W3204348672 hasConcept C152124472 @default.
- W3204348672 hasConcept C153083717 @default.
- W3204348672 hasConcept C154945302 @default.
- W3204348672 hasConcept C166366890 @default.
- W3204348672 hasConcept C200801453 @default.
- W3204348672 hasConcept C28855332 @default.
- W3204348672 hasConcept C41008148 @default.
- W3204348672 hasConcept C50644808 @default.
- W3204348672 hasConcept C57273362 @default.
- W3204348672 hasConcept C62520636 @default.
- W3204348672 hasConcept C78548338 @default.
- W3204348672 hasConcept C79403827 @default.
- W3204348672 hasConceptScore W3204348672C101738243 @default.
- W3204348672 hasConceptScore W3204348672C106301342 @default.
- W3204348672 hasConceptScore W3204348672C111919701 @default.
- W3204348672 hasConceptScore W3204348672C113775141 @default.
- W3204348672 hasConceptScore W3204348672C11413529 @default.
- W3204348672 hasConceptScore W3204348672C118505674 @default.
- W3204348672 hasConceptScore W3204348672C120314980 @default.
- W3204348672 hasConceptScore W3204348672C121332964 @default.
- W3204348672 hasConceptScore W3204348672C152124472 @default.
- W3204348672 hasConceptScore W3204348672C153083717 @default.
- W3204348672 hasConceptScore W3204348672C154945302 @default.
- W3204348672 hasConceptScore W3204348672C166366890 @default.
- W3204348672 hasConceptScore W3204348672C200801453 @default.
- W3204348672 hasConceptScore W3204348672C28855332 @default.
- W3204348672 hasConceptScore W3204348672C41008148 @default.
- W3204348672 hasConceptScore W3204348672C50644808 @default.
- W3204348672 hasConceptScore W3204348672C57273362 @default.
- W3204348672 hasConceptScore W3204348672C62520636 @default.
- W3204348672 hasConceptScore W3204348672C78548338 @default.
- W3204348672 hasConceptScore W3204348672C79403827 @default.
- W3204348672 hasLocation W32043486721 @default.
- W3204348672 hasOpenAccess W3204348672 @default.
- W3204348672 hasPrimaryLocation W32043486721 @default.
- W3204348672 hasRelatedWork W1517259532 @default.
- W3204348672 hasRelatedWork W1730943557 @default.
- W3204348672 hasRelatedWork W2081829887 @default.
- W3204348672 hasRelatedWork W2099678935 @default.
- W3204348672 hasRelatedWork W2123254197 @default.
- W3204348672 hasRelatedWork W2924535634 @default.
- W3204348672 hasRelatedWork W3105229232 @default.
- W3204348672 hasRelatedWork W3148873737 @default.
- W3204348672 hasRelatedWork W4293117873 @default.
- W3204348672 hasRelatedWork W4353114386 @default.
- W3204348672 isParatext "false" @default.
- W3204348672 isRetracted "false" @default.
- W3204348672 magId "3204348672" @default.
- W3204348672 workType "article" @default.