Matches in SemOpenAlex for { <https://semopenalex.org/work/W3112972317> ?p ?o ?g. }
Showing items 1 to 71 of
71
with 100 items per page.
- W3112972317 endingPage "2274" @default.
- W3112972317 startingPage "2261" @default.
- W3112972317 abstract "Federated Learning (FL) serves privacy-preserving collaborative learning among multiple isolated parties, while retaining their privacy data locally. Cross-device and cross-silo FL have achieved great success in cross-domain applications, in which the scarce communication resource is the primary bottleneck. Driven by the need to combine heterogeneous machines from different parties to build a shared data center, we found <italic xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>intra-domain FL</i> , a new type of FL in which isolated parties collaborate in the shared data center, and strong computational heterogeneity becomes the primary bottleneck. To mitigate the training inefficiency caused by stragglers, this article proposes an efficient synchronization algorithm <italic xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>ESync</i> , which allows parties to train different iterations locally under the coordination of a novel scheduler <italic xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>State Server</i> . We give the boundaries of weight divergence and optimality gap of <italic xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>ESync</i> , and analyze the trade-off between convergence accuracy and communication efficiency. Extensive experiments are conducted to compare <italic xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>ESync</i> with SSGD, ASGD, DC-ASGD, FedAvg, FedAsync, TiFL, and FedDrop under strong computational heterogeneity. Numerical results show that <italic xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>ESync</i> achieves great speed up without loss of accuracy, and therefore demonstrate the effectiveness of <italic xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>ESync</i> in both training efficiency and converged accuracy." @default.
- W3112972317 created "2020-12-21" @default.
- W3112972317 creator A5022284072 @default.
- W3112972317 creator A5024438027 @default.
- W3112972317 creator A5039410493 @default.
- W3112972317 creator A5051227924 @default.
- W3112972317 creator A5053945147 @default.
- W3112972317 creator A5060960499 @default.
- W3112972317 date "2022-07-01" @default.
- W3112972317 modified "2023-10-12" @default.
- W3112972317 title "ESync: Accelerating Intra-Domain Federated Learning in Heterogeneous Data Centers" @default.
- W3112972317 cites W2183341477 @default.
- W3112972317 cites W2194775991 @default.
- W3112972317 cites W2744999500 @default.
- W3112972317 cites W2782985982 @default.
- W3112972317 cites W2912213068 @default.
- W3112972317 cites W2936343019 @default.
- W3112972317 cites W3037871107 @default.
- W3112972317 doi "https://doi.org/10.1109/tsc.2020.3044043" @default.
- W3112972317 hasPublicationYear "2022" @default.
- W3112972317 type Work @default.
- W3112972317 sameAs 3112972317 @default.
- W3112972317 citedByCount "8" @default.
- W3112972317 countsByYear W31129723172022 @default.
- W3112972317 countsByYear W31129723172023 @default.
- W3112972317 crossrefType "journal-article" @default.
- W3112972317 hasAuthorship W3112972317A5022284072 @default.
- W3112972317 hasAuthorship W3112972317A5024438027 @default.
- W3112972317 hasAuthorship W3112972317A5039410493 @default.
- W3112972317 hasAuthorship W3112972317A5051227924 @default.
- W3112972317 hasAuthorship W3112972317A5053945147 @default.
- W3112972317 hasAuthorship W3112972317A5060960499 @default.
- W3112972317 hasConcept C119857082 @default.
- W3112972317 hasConcept C134306372 @default.
- W3112972317 hasConcept C149635348 @default.
- W3112972317 hasConcept C154945302 @default.
- W3112972317 hasConcept C2780513914 @default.
- W3112972317 hasConcept C33923547 @default.
- W3112972317 hasConcept C36503486 @default.
- W3112972317 hasConcept C41008148 @default.
- W3112972317 hasConcept C80444323 @default.
- W3112972317 hasConceptScore W3112972317C119857082 @default.
- W3112972317 hasConceptScore W3112972317C134306372 @default.
- W3112972317 hasConceptScore W3112972317C149635348 @default.
- W3112972317 hasConceptScore W3112972317C154945302 @default.
- W3112972317 hasConceptScore W3112972317C2780513914 @default.
- W3112972317 hasConceptScore W3112972317C33923547 @default.
- W3112972317 hasConceptScore W3112972317C36503486 @default.
- W3112972317 hasConceptScore W3112972317C41008148 @default.
- W3112972317 hasConceptScore W3112972317C80444323 @default.
- W3112972317 hasIssue "4" @default.
- W3112972317 hasLocation W31129723171 @default.
- W3112972317 hasOpenAccess W3112972317 @default.
- W3112972317 hasPrimaryLocation W31129723171 @default.
- W3112972317 hasRelatedWork W2961085424 @default.
- W3112972317 hasRelatedWork W3046775127 @default.
- W3112972317 hasRelatedWork W3107602296 @default.
- W3112972317 hasRelatedWork W3170094116 @default.
- W3112972317 hasRelatedWork W3209574120 @default.
- W3112972317 hasRelatedWork W4210805261 @default.
- W3112972317 hasRelatedWork W4306674287 @default.
- W3112972317 hasRelatedWork W4312192474 @default.
- W3112972317 hasRelatedWork W4386462264 @default.
- W3112972317 hasRelatedWork W4387297750 @default.
- W3112972317 hasVolume "15" @default.
- W3112972317 isParatext "false" @default.
- W3112972317 isRetracted "false" @default.
- W3112972317 magId "3112972317" @default.
- W3112972317 workType "article" @default.