Matches in SemOpenAlex for { <https://semopenalex.org/work/W4319862449> ?p ?o ?g. }
Showing items 1 to 80 of
80
with 100 items per page.
- W4319862449 abstract "Machine learning model weights and activations are represented in full-precision during training. This leads to performance degradation in runtime when deployed on neural network accelerator (NNA) chips, which leverage highly parallelized fixed-point arithmetic to improve runtime memory and latency. In this work, we replicate the NNA operators during the training phase, accounting for the degradation due to low-precision inference on the NNA in back-propagation. Our proposed method efficiently emulates NNA operations, thus foregoing the need to transfer quantization error-prone data to the Central Processing Unit (CPU), ultimately reducing the user perceived latency (UPL). We apply our approach to Recurrent Neural Network-Transducer (RNN-T), an attractive architecture for on-device streaming speech recognition tasks. We train and evaluate models on 270K hours of English data and show a 5-7% improvement in engine latency while saving up to 10% relative degradation in WER." @default.
- W4319862449 created "2023-02-11" @default.
- W4319862449 creator A5008201430 @default.
- W4319862449 creator A5023708721 @default.
- W4319862449 creator A5029573067 @default.
- W4319862449 creator A5038985606 @default.
- W4319862449 creator A5042678478 @default.
- W4319862449 creator A5062762746 @default.
- W4319862449 creator A5066216178 @default.
- W4319862449 creator A5076547892 @default.
- W4319862449 creator A5076634486 @default.
- W4319862449 date "2023-01-09" @default.
- W4319862449 modified "2023-09-29" @default.
- W4319862449 title "Accelerator-Aware Training for Transducer-Based Speech Recognition" @default.
- W4319862449 cites W1689711448 @default.
- W4319862449 cites W2064675550 @default.
- W4319862449 cites W2142343760 @default.
- W4319862449 cites W2167425979 @default.
- W4319862449 cites W2892009249 @default.
- W4319862449 cites W2981857663 @default.
- W4319862449 cites W3151287998 @default.
- W4319862449 cites W3162649911 @default.
- W4319862449 cites W3163062345 @default.
- W4319862449 cites W3196783077 @default.
- W4319862449 cites W3197201410 @default.
- W4319862449 cites W4297841536 @default.
- W4319862449 doi "https://doi.org/10.1109/slt54892.2023.10022592" @default.
- W4319862449 hasPublicationYear "2023" @default.
- W4319862449 type Work @default.
- W4319862449 citedByCount "0" @default.
- W4319862449 crossrefType "proceedings-article" @default.
- W4319862449 hasAuthorship W4319862449A5008201430 @default.
- W4319862449 hasAuthorship W4319862449A5023708721 @default.
- W4319862449 hasAuthorship W4319862449A5029573067 @default.
- W4319862449 hasAuthorship W4319862449A5038985606 @default.
- W4319862449 hasAuthorship W4319862449A5042678478 @default.
- W4319862449 hasAuthorship W4319862449A5062762746 @default.
- W4319862449 hasAuthorship W4319862449A5066216178 @default.
- W4319862449 hasAuthorship W4319862449A5076547892 @default.
- W4319862449 hasAuthorship W4319862449A5076634486 @default.
- W4319862449 hasBestOaLocation W43198624492 @default.
- W4319862449 hasConcept C147168706 @default.
- W4319862449 hasConcept C153083717 @default.
- W4319862449 hasConcept C154945302 @default.
- W4319862449 hasConcept C2776214188 @default.
- W4319862449 hasConcept C28490314 @default.
- W4319862449 hasConcept C28855332 @default.
- W4319862449 hasConcept C31972630 @default.
- W4319862449 hasConcept C41008148 @default.
- W4319862449 hasConcept C50644808 @default.
- W4319862449 hasConcept C76155785 @default.
- W4319862449 hasConcept C82876162 @default.
- W4319862449 hasConceptScore W4319862449C147168706 @default.
- W4319862449 hasConceptScore W4319862449C153083717 @default.
- W4319862449 hasConceptScore W4319862449C154945302 @default.
- W4319862449 hasConceptScore W4319862449C2776214188 @default.
- W4319862449 hasConceptScore W4319862449C28490314 @default.
- W4319862449 hasConceptScore W4319862449C28855332 @default.
- W4319862449 hasConceptScore W4319862449C31972630 @default.
- W4319862449 hasConceptScore W4319862449C41008148 @default.
- W4319862449 hasConceptScore W4319862449C50644808 @default.
- W4319862449 hasConceptScore W4319862449C76155785 @default.
- W4319862449 hasConceptScore W4319862449C82876162 @default.
- W4319862449 hasLocation W43198624491 @default.
- W4319862449 hasLocation W43198624492 @default.
- W4319862449 hasOpenAccess W4319862449 @default.
- W4319862449 hasPrimaryLocation W43198624491 @default.
- W4319862449 hasRelatedWork W2329734087 @default.
- W4319862449 hasRelatedWork W2798291715 @default.
- W4319862449 hasRelatedWork W2987019345 @default.
- W4319862449 hasRelatedWork W3170939428 @default.
- W4319862449 hasRelatedWork W3175075966 @default.
- W4319862449 hasRelatedWork W3184187848 @default.
- W4319862449 hasRelatedWork W3197304116 @default.
- W4319862449 hasRelatedWork W4294982680 @default.
- W4319862449 hasRelatedWork W4312964860 @default.
- W4319862449 hasRelatedWork W4372272185 @default.
- W4319862449 isParatext "false" @default.
- W4319862449 isRetracted "false" @default.
- W4319862449 workType "article" @default.