Matches in SemOpenAlex for { <https://semopenalex.org/work/W2912070033> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W2912070033 abstract "Deep Neural Networks (DNNs) are one of the leading classification algorithms and have achieved big milestones such as GoogleNet and AlphaGo. Training of neural network is a time-consuming affair and is proportional to the depth of network and number of computations carried out in each layer. If the network is deep, the time taken is considerably high as it works sequentially on batches of dataset using Sequential Back-propagation (BP) Algorithm. This paper presents an acceleration technique DAPP Accelerating Training of DNN using Ping-Pong approach to reduce the training time using distributed local memory. DAPP exploits model parallelism without compromising the accuracy of the model. The proposed pipeline design concurrently works on consecutive set of batches to compute gradients at various layers, while eliminating the need for global memory. Simulation of DAPP is done on MNIST and CIFAR-10 datasets using System-C. Additionally, this technique has been adapted for multi-core architectures. This generic methodology has been implemented for CNN, Vanilla RNN and LSTM networks. This design reduces training time by 40% for a 3 layer CNN, 92% for a 10 layer CNN, 38% for Vanilla RNN, and 40% for LSTM while maintaining the accuracy in all above networks." @default.
- W2912070033 created "2019-02-21" @default.
- W2912070033 creator A5033158594 @default.
- W2912070033 creator A5036667823 @default.
- W2912070033 creator A5085835894 @default.
- W2912070033 date "2018-06-01" @default.
- W2912070033 modified "2023-09-23" @default.
- W2912070033 title "DAPP: Accelerating Training of DNN" @default.
- W2912070033 cites W2087402357 @default.
- W2912070033 cites W2336650964 @default.
- W2912070033 cites W2552496680 @default.
- W2912070033 cites W2592417500 @default.
- W2912070033 cites W2741430497 @default.
- W2912070033 cites W2963804082 @default.
- W2912070033 cites W3118608800 @default.
- W2912070033 cites W3142400242 @default.
- W2912070033 doi "https://doi.org/10.1109/hpcc/smartcity/dss.2018.00144" @default.
- W2912070033 hasPublicationYear "2018" @default.
- W2912070033 type Work @default.
- W2912070033 sameAs 2912070033 @default.
- W2912070033 citedByCount "0" @default.
- W2912070033 crossrefType "proceedings-article" @default.
- W2912070033 hasAuthorship W2912070033A5033158594 @default.
- W2912070033 hasAuthorship W2912070033A5036667823 @default.
- W2912070033 hasAuthorship W2912070033A5085835894 @default.
- W2912070033 hasConcept C108583219 @default.
- W2912070033 hasConcept C11413529 @default.
- W2912070033 hasConcept C117896860 @default.
- W2912070033 hasConcept C119857082 @default.
- W2912070033 hasConcept C121332964 @default.
- W2912070033 hasConcept C147168706 @default.
- W2912070033 hasConcept C154945302 @default.
- W2912070033 hasConcept C165696696 @default.
- W2912070033 hasConcept C177264268 @default.
- W2912070033 hasConcept C178790620 @default.
- W2912070033 hasConcept C185592680 @default.
- W2912070033 hasConcept C190502265 @default.
- W2912070033 hasConcept C199360897 @default.
- W2912070033 hasConcept C2779227376 @default.
- W2912070033 hasConcept C38652104 @default.
- W2912070033 hasConcept C41008148 @default.
- W2912070033 hasConcept C43521106 @default.
- W2912070033 hasConcept C45374587 @default.
- W2912070033 hasConcept C50644808 @default.
- W2912070033 hasConcept C74650414 @default.
- W2912070033 hasConcept C81363708 @default.
- W2912070033 hasConceptScore W2912070033C108583219 @default.
- W2912070033 hasConceptScore W2912070033C11413529 @default.
- W2912070033 hasConceptScore W2912070033C117896860 @default.
- W2912070033 hasConceptScore W2912070033C119857082 @default.
- W2912070033 hasConceptScore W2912070033C121332964 @default.
- W2912070033 hasConceptScore W2912070033C147168706 @default.
- W2912070033 hasConceptScore W2912070033C154945302 @default.
- W2912070033 hasConceptScore W2912070033C165696696 @default.
- W2912070033 hasConceptScore W2912070033C177264268 @default.
- W2912070033 hasConceptScore W2912070033C178790620 @default.
- W2912070033 hasConceptScore W2912070033C185592680 @default.
- W2912070033 hasConceptScore W2912070033C190502265 @default.
- W2912070033 hasConceptScore W2912070033C199360897 @default.
- W2912070033 hasConceptScore W2912070033C2779227376 @default.
- W2912070033 hasConceptScore W2912070033C38652104 @default.
- W2912070033 hasConceptScore W2912070033C41008148 @default.
- W2912070033 hasConceptScore W2912070033C43521106 @default.
- W2912070033 hasConceptScore W2912070033C45374587 @default.
- W2912070033 hasConceptScore W2912070033C50644808 @default.
- W2912070033 hasConceptScore W2912070033C74650414 @default.
- W2912070033 hasConceptScore W2912070033C81363708 @default.
- W2912070033 hasLocation W29120700331 @default.
- W2912070033 hasOpenAccess W2912070033 @default.
- W2912070033 hasPrimaryLocation W29120700331 @default.
- W2912070033 hasRelatedWork W1639373162 @default.
- W2912070033 hasRelatedWork W2758217051 @default.
- W2912070033 hasRelatedWork W2794361067 @default.
- W2912070033 hasRelatedWork W2796313706 @default.
- W2912070033 hasRelatedWork W2799137784 @default.
- W2912070033 hasRelatedWork W2896053637 @default.
- W2912070033 hasRelatedWork W2898078303 @default.
- W2912070033 hasRelatedWork W2899742314 @default.
- W2912070033 hasRelatedWork W2909175897 @default.
- W2912070033 hasRelatedWork W2913263636 @default.
- W2912070033 hasRelatedWork W2931633209 @default.
- W2912070033 hasRelatedWork W2976097345 @default.
- W2912070033 hasRelatedWork W3016118517 @default.
- W2912070033 hasRelatedWork W3016829100 @default.
- W2912070033 hasRelatedWork W3120936489 @default.
- W2912070033 hasRelatedWork W3121874409 @default.
- W2912070033 hasRelatedWork W3159522174 @default.
- W2912070033 hasRelatedWork W3160382941 @default.
- W2912070033 hasRelatedWork W3195381335 @default.
- W2912070033 hasRelatedWork W326710513 @default.
- W2912070033 isParatext "false" @default.
- W2912070033 isRetracted "false" @default.
- W2912070033 magId "2912070033" @default.
- W2912070033 workType "article" @default.