Matches in SemOpenAlex for { <https://semopenalex.org/work/W3044893339> ?p ?o ?g. }
- W3044893339 abstract "The scalability of Distributed Stochastic Gradient Descent (SGD) is today limited by communication bottlenecks. We propose a novel SGD variant: Communication-efficient SGD with Error Reset, or CSER. The key idea in CSER is first a new technique called error reset that adapts arbitrary compressors for SGD, producing bifurcated local models with periodic reset of resulting local residual errors. Second we introduce partial synchronization for both the gradients and the models, leveraging advantages from them. We prove the convergence of CSER for smooth non-convex problems. Empirical results show that when combined with highly aggressive compressors, the CSER algorithms accelerate the distributed training by nearly 10x for CIFAR-100, and by 4.5x for ImageNet." @default.
- W3044893339 created "2020-07-29" @default.
- W3044893339 creator A5027240302 @default.
- W3044893339 creator A5042419001 @default.
- W3044893339 creator A5051065845 @default.
- W3044893339 creator A5069243328 @default.
- W3044893339 creator A5076316802 @default.
- W3044893339 creator A5084047386 @default.
- W3044893339 date "2020-07-26" @default.
- W3044893339 modified "2023-09-25" @default.
- W3044893339 title "CSER: Communication-efficient SGD with Error Reset" @default.
- W3044893339 cites W104184427 @default.
- W3044893339 cites W2083842231 @default.
- W3044893339 cites W2117539524 @default.
- W3044893339 cites W2127941149 @default.
- W3044893339 cites W2132737349 @default.
- W3044893339 cites W2302255633 @default.
- W3044893339 cites W2405578611 @default.
- W3044893339 cites W2407022425 @default.
- W3044893339 cites W2622263826 @default.
- W3044893339 cites W2749988060 @default.
- W3044893339 cites W2769644379 @default.
- W3044893339 cites W2787998955 @default.
- W3044893339 cites W2888561381 @default.
- W3044893339 cites W2890924858 @default.
- W3044893339 cites W2891735383 @default.
- W3044893339 cites W2900182564 @default.
- W3044893339 cites W2949934631 @default.
- W3044893339 cites W2950826569 @default.
- W3044893339 cites W2962747323 @default.
- W3044893339 cites W2963179579 @default.
- W3044893339 cites W2963263347 @default.
- W3044893339 cites W2963664311 @default.
- W3044893339 cites W2964004663 @default.
- W3044893339 cites W2964137095 @default.
- W3044893339 cites W2970289928 @default.
- W3044893339 cites W2971064744 @default.
- W3044893339 cites W2971342441 @default.
- W3044893339 cites W2994779554 @default.
- W3044893339 cites W3101036738 @default.
- W3044893339 cites W3118608800 @default.
- W3044893339 doi "https://doi.org/10.48550/arxiv.2007.13221" @default.
- W3044893339 hasPublicationYear "2020" @default.
- W3044893339 type Work @default.
- W3044893339 sameAs 3044893339 @default.
- W3044893339 citedByCount "1" @default.
- W3044893339 countsByYear W30448933392021 @default.
- W3044893339 crossrefType "posted-content" @default.
- W3044893339 hasAuthorship W3044893339A5027240302 @default.
- W3044893339 hasAuthorship W3044893339A5042419001 @default.
- W3044893339 hasAuthorship W3044893339A5051065845 @default.
- W3044893339 hasAuthorship W3044893339A5069243328 @default.
- W3044893339 hasAuthorship W3044893339A5076316802 @default.
- W3044893339 hasAuthorship W3044893339A5084047386 @default.
- W3044893339 hasBestOaLocation W30448933391 @default.
- W3044893339 hasConcept C106159729 @default.
- W3044893339 hasConcept C11413529 @default.
- W3044893339 hasConcept C120314980 @default.
- W3044893339 hasConcept C127162648 @default.
- W3044893339 hasConcept C154945302 @default.
- W3044893339 hasConcept C155512373 @default.
- W3044893339 hasConcept C162324750 @default.
- W3044893339 hasConcept C206688291 @default.
- W3044893339 hasConcept C2777303404 @default.
- W3044893339 hasConcept C2778562939 @default.
- W3044893339 hasConcept C2779795794 @default.
- W3044893339 hasConcept C41008148 @default.
- W3044893339 hasConcept C48044578 @default.
- W3044893339 hasConcept C50522688 @default.
- W3044893339 hasConcept C50644808 @default.
- W3044893339 hasConcept C76155785 @default.
- W3044893339 hasConcept C77088390 @default.
- W3044893339 hasConcept C80444323 @default.
- W3044893339 hasConceptScore W3044893339C106159729 @default.
- W3044893339 hasConceptScore W3044893339C11413529 @default.
- W3044893339 hasConceptScore W3044893339C120314980 @default.
- W3044893339 hasConceptScore W3044893339C127162648 @default.
- W3044893339 hasConceptScore W3044893339C154945302 @default.
- W3044893339 hasConceptScore W3044893339C155512373 @default.
- W3044893339 hasConceptScore W3044893339C162324750 @default.
- W3044893339 hasConceptScore W3044893339C206688291 @default.
- W3044893339 hasConceptScore W3044893339C2777303404 @default.
- W3044893339 hasConceptScore W3044893339C2778562939 @default.
- W3044893339 hasConceptScore W3044893339C2779795794 @default.
- W3044893339 hasConceptScore W3044893339C41008148 @default.
- W3044893339 hasConceptScore W3044893339C48044578 @default.
- W3044893339 hasConceptScore W3044893339C50522688 @default.
- W3044893339 hasConceptScore W3044893339C50644808 @default.
- W3044893339 hasConceptScore W3044893339C76155785 @default.
- W3044893339 hasConceptScore W3044893339C77088390 @default.
- W3044893339 hasConceptScore W3044893339C80444323 @default.
- W3044893339 hasLocation W30448933391 @default.
- W3044893339 hasOpenAccess W3044893339 @default.
- W3044893339 hasPrimaryLocation W30448933391 @default.
- W3044893339 hasRelatedWork W1588015694 @default.
- W3044893339 hasRelatedWork W1596201972 @default.
- W3044893339 hasRelatedWork W1767718647 @default.
- W3044893339 hasRelatedWork W1788737569 @default.
- W3044893339 hasRelatedWork W1967954938 @default.
- W3044893339 hasRelatedWork W1986253068 @default.