Matches in SemOpenAlex for { <https://semopenalex.org/work/W2952857865> ?p ?o ?g. }
- W2952857865 abstract "Training neural networks with back-propagation (BP) requires a sequential passing of activations and gradients, which forces the network modules to work in a synchronous fashion. This has been recognized as the lockings (i.e., the forward, backward and update lockings) inherited from the BP. In this paper, we propose a fully decoupled training scheme using delayed gradients (FDG) to break all these lockings. The FDG splits a neural network into multiple modules and trains them independently and asynchronously using different workers (e.g., GPUs). We also introduce a gradient shrinking process to reduce the stale gradient effect caused by the delayed gradients. In addition, we prove that the proposed FDG algorithm guarantees a statistical convergence during training. Experiments are conducted by training deep convolutional neural networks to perform classification tasks on benchmark datasets, showing comparable or better results against the state-of-the-art methods as well as the BP in terms of both generalization and acceleration abilities. In particular, we show that the FDG is also able to train very wide networks (e.g., WRN-28-10) and extremely deep networks (e.g., ResNet-1202). Code is available at https://github.com/ZHUANGHP/FDG." @default.
- W2952857865 created "2019-06-27" @default.
- W2952857865 creator A5007322337 @default.
- W2952857865 creator A5028602831 @default.
- W2952857865 creator A5049506273 @default.
- W2952857865 creator A5061256037 @default.
- W2952857865 creator A5062005306 @default.
- W2952857865 date "2019-06-21" @default.
- W2952857865 modified "2023-10-16" @default.
- W2952857865 title "Fully Decoupled Neural Network Learning Using Delayed Gradients" @default.
- W2952857865 cites W2064675550 @default.
- W2952857865 cites W2112796928 @default.
- W2952857865 cites W2112952404 @default.
- W2952857865 cites W2157331557 @default.
- W2952857865 cites W2168231600 @default.
- W2952857865 cites W2194775991 @default.
- W2952857865 cites W2336650964 @default.
- W2952857865 cites W2401231614 @default.
- W2952857865 cites W2516591743 @default.
- W2952857865 cites W2549139847 @default.
- W2952857865 cites W2552737632 @default.
- W2952857865 cites W2612387305 @default.
- W2952857865 cites W2619184049 @default.
- W2952857865 cites W2786738752 @default.
- W2952857865 cites W2787998955 @default.
- W2952857865 cites W2884700152 @default.
- W2952857865 cites W2911586496 @default.
- W2952857865 cites W2912855321 @default.
- W2952857865 cites W2949117887 @default.
- W2952857865 cites W2963373778 @default.
- W2952857865 cites W2963446712 @default.
- W2952857865 cites W2963903325 @default.
- W2952857865 cites W2964113612 @default.
- W2952857865 cites W2964115671 @default.
- W2952857865 cites W2964319207 @default.
- W2952857865 cites W2991040477 @default.
- W2952857865 cites W3118608800 @default.
- W2952857865 cites W3121926921 @default.
- W2952857865 doi "https://doi.org/10.48550/arxiv.1906.09108" @default.
- W2952857865 hasPublicationYear "2019" @default.
- W2952857865 type Work @default.
- W2952857865 sameAs 2952857865 @default.
- W2952857865 citedByCount "6" @default.
- W2952857865 countsByYear W29528578652020 @default.
- W2952857865 countsByYear W29528578652021 @default.
- W2952857865 crossrefType "posted-content" @default.
- W2952857865 hasAuthorship W2952857865A5007322337 @default.
- W2952857865 hasAuthorship W2952857865A5028602831 @default.
- W2952857865 hasAuthorship W2952857865A5049506273 @default.
- W2952857865 hasAuthorship W2952857865A5061256037 @default.
- W2952857865 hasAuthorship W2952857865A5062005306 @default.
- W2952857865 hasBestOaLocation W29528578651 @default.
- W2952857865 hasConcept C108583219 @default.
- W2952857865 hasConcept C111919701 @default.
- W2952857865 hasConcept C11413529 @default.
- W2952857865 hasConcept C117896860 @default.
- W2952857865 hasConcept C121332964 @default.
- W2952857865 hasConcept C13280743 @default.
- W2952857865 hasConcept C134306372 @default.
- W2952857865 hasConcept C153258448 @default.
- W2952857865 hasConcept C154945302 @default.
- W2952857865 hasConcept C155032097 @default.
- W2952857865 hasConcept C162324750 @default.
- W2952857865 hasConcept C177148314 @default.
- W2952857865 hasConcept C177264268 @default.
- W2952857865 hasConcept C185798385 @default.
- W2952857865 hasConcept C190839683 @default.
- W2952857865 hasConcept C199360897 @default.
- W2952857865 hasConcept C205649164 @default.
- W2952857865 hasConcept C2776760102 @default.
- W2952857865 hasConcept C2777303404 @default.
- W2952857865 hasConcept C2984842247 @default.
- W2952857865 hasConcept C33923547 @default.
- W2952857865 hasConcept C41008148 @default.
- W2952857865 hasConcept C50522688 @default.
- W2952857865 hasConcept C50644808 @default.
- W2952857865 hasConcept C58640448 @default.
- W2952857865 hasConcept C74650414 @default.
- W2952857865 hasConcept C81363708 @default.
- W2952857865 hasConcept C98045186 @default.
- W2952857865 hasConceptScore W2952857865C108583219 @default.
- W2952857865 hasConceptScore W2952857865C111919701 @default.
- W2952857865 hasConceptScore W2952857865C11413529 @default.
- W2952857865 hasConceptScore W2952857865C117896860 @default.
- W2952857865 hasConceptScore W2952857865C121332964 @default.
- W2952857865 hasConceptScore W2952857865C13280743 @default.
- W2952857865 hasConceptScore W2952857865C134306372 @default.
- W2952857865 hasConceptScore W2952857865C153258448 @default.
- W2952857865 hasConceptScore W2952857865C154945302 @default.
- W2952857865 hasConceptScore W2952857865C155032097 @default.
- W2952857865 hasConceptScore W2952857865C162324750 @default.
- W2952857865 hasConceptScore W2952857865C177148314 @default.
- W2952857865 hasConceptScore W2952857865C177264268 @default.
- W2952857865 hasConceptScore W2952857865C185798385 @default.
- W2952857865 hasConceptScore W2952857865C190839683 @default.
- W2952857865 hasConceptScore W2952857865C199360897 @default.
- W2952857865 hasConceptScore W2952857865C205649164 @default.
- W2952857865 hasConceptScore W2952857865C2776760102 @default.
- W2952857865 hasConceptScore W2952857865C2777303404 @default.
- W2952857865 hasConceptScore W2952857865C2984842247 @default.