Matches in SemOpenAlex for { <https://semopenalex.org/work/W4210961870> ?p ?o ?g. }
- W4210961870 endingPage "24" @default.
- W4210961870 startingPage "1" @default.
- W4210961870 abstract "With the prospering of mobile devices, the distributed learning approach, enabling model training with decentralized data, has attracted great interest from researchers. However, the lack of training capability for edge devices significantly limits the energy efficiency of distributed learning in real life. This article describes Efficient-Grad, an algorithm-hardware co-design approach for training deep convolutional neural networks, which improves both throughput and energy saving during model training, with negligible validation accuracy loss. The key to Efficient-Grad is its exploitation of two observations. Firstly, the sparsity has potential for not only activation and weight, but gradients and the asymmetry residing in the gradients for the conventional back propagation (BP). Secondly, a dedicated hardware architecture for sparsity utilization and efficient data movement can be optimized to support the Efficient-Grad algorithm in a scalable manner. To the best of our knowledge, Efficient-Grad is the first approach that successfully adopts a feedback-alignment (FA)-based gradient optimization scheme for deep convolutional neural network training, which leads to its superiority in terms of energy efficiency. We present case studies to demonstrate that the Efficient-Grad design outperforms the prior arts by 3.72x in terms of energy efficiency." @default.
- W4210961870 created "2022-02-13" @default.
- W4210961870 creator A5004527275 @default.
- W4210961870 creator A5052377688 @default.
- W4210961870 date "2022-02-08" @default.
- W4210961870 modified "2023-09-25" @default.
- W4210961870 title "Efficient-Grad: <u>Efficient</u> Training Deep Convolutional Neural Networks on Edge Devices with <u>Grad</u> ient Optimizations" @default.
- W4210961870 cites W1498436455 @default.
- W4210961870 cites W1983394510 @default.
- W4210961870 cites W1999085092 @default.
- W4210961870 cites W2048266589 @default.
- W4210961870 cites W2058539177 @default.
- W4210961870 cites W2067523571 @default.
- W4210961870 cites W2108598243 @default.
- W4210961870 cites W2152839228 @default.
- W4210961870 cites W2162639668 @default.
- W4210961870 cites W2171318521 @default.
- W4210961870 cites W2194775991 @default.
- W4210961870 cites W2552737632 @default.
- W4210961870 cites W2605487586 @default.
- W4210961870 cites W2606722458 @default.
- W4210961870 cites W2611105550 @default.
- W4210961870 cites W2773339846 @default.
- W4210961870 cites W2774669597 @default.
- W4210961870 cites W2898985762 @default.
- W4210961870 cites W2902138814 @default.
- W4210961870 cites W2909119455 @default.
- W4210961870 cites W2912213068 @default.
- W4210961870 cites W2921918777 @default.
- W4210961870 cites W2944850220 @default.
- W4210961870 cites W2945146780 @default.
- W4210961870 cites W2945992628 @default.
- W4210961870 cites W2964787841 @default.
- W4210961870 cites W3006586535 @default.
- W4210961870 cites W3006769269 @default.
- W4210961870 cites W3015764133 @default.
- W4210961870 cites W3027968530 @default.
- W4210961870 cites W3040850704 @default.
- W4210961870 cites W3090288569 @default.
- W4210961870 cites W3091629465 @default.
- W4210961870 cites W3092357178 @default.
- W4210961870 cites W3097528158 @default.
- W4210961870 cites W3117204317 @default.
- W4210961870 cites W3121174466 @default.
- W4210961870 cites W3135859967 @default.
- W4210961870 cites W3148444620 @default.
- W4210961870 cites W3213528054 @default.
- W4210961870 doi "https://doi.org/10.1145/3504034" @default.
- W4210961870 hasPublicationYear "2022" @default.
- W4210961870 type Work @default.
- W4210961870 citedByCount "1" @default.
- W4210961870 countsByYear W42109618702023 @default.
- W4210961870 crossrefType "journal-article" @default.
- W4210961870 hasAuthorship W4210961870A5004527275 @default.
- W4210961870 hasAuthorship W4210961870A5052377688 @default.
- W4210961870 hasConcept C108583219 @default.
- W4210961870 hasConcept C111919701 @default.
- W4210961870 hasConcept C113775141 @default.
- W4210961870 hasConcept C119599485 @default.
- W4210961870 hasConcept C119857082 @default.
- W4210961870 hasConcept C121332964 @default.
- W4210961870 hasConcept C127413603 @default.
- W4210961870 hasConcept C134306372 @default.
- W4210961870 hasConcept C138236772 @default.
- W4210961870 hasConcept C153294291 @default.
- W4210961870 hasConcept C154945302 @default.
- W4210961870 hasConcept C157764524 @default.
- W4210961870 hasConcept C162307627 @default.
- W4210961870 hasConcept C2742236 @default.
- W4210961870 hasConcept C2777211547 @default.
- W4210961870 hasConcept C2984842247 @default.
- W4210961870 hasConcept C33923547 @default.
- W4210961870 hasConcept C41008148 @default.
- W4210961870 hasConcept C48044578 @default.
- W4210961870 hasConcept C50644808 @default.
- W4210961870 hasConcept C555944384 @default.
- W4210961870 hasConcept C76155785 @default.
- W4210961870 hasConcept C77088390 @default.
- W4210961870 hasConcept C77618280 @default.
- W4210961870 hasConcept C79974875 @default.
- W4210961870 hasConcept C81363708 @default.
- W4210961870 hasConceptScore W4210961870C108583219 @default.
- W4210961870 hasConceptScore W4210961870C111919701 @default.
- W4210961870 hasConceptScore W4210961870C113775141 @default.
- W4210961870 hasConceptScore W4210961870C119599485 @default.
- W4210961870 hasConceptScore W4210961870C119857082 @default.
- W4210961870 hasConceptScore W4210961870C121332964 @default.
- W4210961870 hasConceptScore W4210961870C127413603 @default.
- W4210961870 hasConceptScore W4210961870C134306372 @default.
- W4210961870 hasConceptScore W4210961870C138236772 @default.
- W4210961870 hasConceptScore W4210961870C153294291 @default.
- W4210961870 hasConceptScore W4210961870C154945302 @default.
- W4210961870 hasConceptScore W4210961870C157764524 @default.
- W4210961870 hasConceptScore W4210961870C162307627 @default.
- W4210961870 hasConceptScore W4210961870C2742236 @default.
- W4210961870 hasConceptScore W4210961870C2777211547 @default.
- W4210961870 hasConceptScore W4210961870C2984842247 @default.
- W4210961870 hasConceptScore W4210961870C33923547 @default.