Matches in SemOpenAlex for { <https://semopenalex.org/work/W3140941716> ?p ?o ?g. }
Showing items 1 to 92 of
92
with 100 items per page.
- W3140941716 abstract "Neural architecture search (NAS) has been successfully applied to tasks like image classification and language modeling for finding efficient high-performance network architectures. In ASR field especially end-to-end ASR, the related research is still in its infancy. In this work, we focus on applying NAS on the most popular manually designed model: Conformer, and then propose an efficient ASR model searching method that benefits from the natural advantage of differentiable architecture search (Darts) in reducing computational overheads. We fuse Darts mutator and Conformer blocks to form a complete search space, within which a modified architecture called Darts-Conformer cell is found automatically. The entire searching process on AISHELL-1 dataset costs only 0.7 GPU days. Replacing the Conformer encoder by stacking searched cell, we get an end-to-end ASR model (named as Darts-Conformner) that outperforms the Conformer baseline by 4.7% on the open-source AISHELL-1 dataset. Besides, we verify the transferability of the architecture searched on a small dataset to a larger 2k-hour dataset. To the best of our knowledge, this is the first successful attempt to apply gradient-based architecture search in the attention-based encoder-decoder ASR model." @default.
- W3140941716 created "2021-04-13" @default.
- W3140941716 creator A5044307720 @default.
- W3140941716 creator A5045685496 @default.
- W3140941716 creator A5057060591 @default.
- W3140941716 creator A5066245750 @default.
- W3140941716 date "2021-04-06" @default.
- W3140941716 modified "2023-10-16" @default.
- W3140941716 title "Darts-Conformer: Towards Efficient Gradient-Based Neural Architecture Search For End-to-End ASR" @default.
- W3140941716 cites W2127141656 @default.
- W3140941716 cites W2327501763 @default.
- W3140941716 cites W2526425061 @default.
- W3140941716 cites W2553303224 @default.
- W3140941716 cites W2594529350 @default.
- W3140941716 cites W2626778328 @default.
- W3140941716 cites W2905672847 @default.
- W3140941716 cites W2939297570 @default.
- W3140941716 cites W2941814890 @default.
- W3140941716 cites W2948981900 @default.
- W3140941716 cites W2951104886 @default.
- W3140941716 cites W2951886768 @default.
- W3140941716 cites W2962780374 @default.
- W3140941716 cites W2963032572 @default.
- W3140941716 cites W2963242190 @default.
- W3140941716 cites W2963414781 @default.
- W3140941716 cites W2964081807 @default.
- W3140941716 cites W2964212578 @default.
- W3140941716 cites W2964259004 @default.
- W3140941716 cites W2965658867 @default.
- W3140941716 cites W2972482531 @default.
- W3140941716 cites W2972830778 @default.
- W3140941716 cites W3007227084 @default.
- W3140941716 cites W3025165719 @default.
- W3140941716 cites W3033604713 @default.
- W3140941716 cites W3043474396 @default.
- W3140941716 cites W3081497373 @default.
- W3140941716 cites W3084006547 @default.
- W3140941716 cites W3096329809 @default.
- W3140941716 cites W3101755305 @default.
- W3140941716 cites W3121327306 @default.
- W3140941716 cites W3127424476 @default.
- W3140941716 doi "https://doi.org/10.48550/arxiv.2104.02868" @default.
- W3140941716 hasPublicationYear "2021" @default.
- W3140941716 type Work @default.
- W3140941716 sameAs 3140941716 @default.
- W3140941716 citedByCount "1" @default.
- W3140941716 countsByYear W31409417162021 @default.
- W3140941716 crossrefType "posted-content" @default.
- W3140941716 hasAuthorship W3140941716A5044307720 @default.
- W3140941716 hasAuthorship W3140941716A5045685496 @default.
- W3140941716 hasAuthorship W3140941716A5057060591 @default.
- W3140941716 hasAuthorship W3140941716A5066245750 @default.
- W3140941716 hasBestOaLocation W31409417161 @default.
- W3140941716 hasConcept C111919701 @default.
- W3140941716 hasConcept C118505674 @default.
- W3140941716 hasConcept C123657996 @default.
- W3140941716 hasConcept C142362112 @default.
- W3140941716 hasConcept C153349607 @default.
- W3140941716 hasConcept C154945302 @default.
- W3140941716 hasConcept C178790620 @default.
- W3140941716 hasConcept C185592680 @default.
- W3140941716 hasConcept C18705241 @default.
- W3140941716 hasConcept C32909587 @default.
- W3140941716 hasConcept C41008148 @default.
- W3140941716 hasConceptScore W3140941716C111919701 @default.
- W3140941716 hasConceptScore W3140941716C118505674 @default.
- W3140941716 hasConceptScore W3140941716C123657996 @default.
- W3140941716 hasConceptScore W3140941716C142362112 @default.
- W3140941716 hasConceptScore W3140941716C153349607 @default.
- W3140941716 hasConceptScore W3140941716C154945302 @default.
- W3140941716 hasConceptScore W3140941716C178790620 @default.
- W3140941716 hasConceptScore W3140941716C185592680 @default.
- W3140941716 hasConceptScore W3140941716C18705241 @default.
- W3140941716 hasConceptScore W3140941716C32909587 @default.
- W3140941716 hasConceptScore W3140941716C41008148 @default.
- W3140941716 hasLocation W31409417161 @default.
- W3140941716 hasOpenAccess W3140941716 @default.
- W3140941716 hasPrimaryLocation W31409417161 @default.
- W3140941716 hasRelatedWork W155493962 @default.
- W3140941716 hasRelatedWork W2358489738 @default.
- W3140941716 hasRelatedWork W2382623646 @default.
- W3140941716 hasRelatedWork W2385621972 @default.
- W3140941716 hasRelatedWork W3107474891 @default.
- W3140941716 hasRelatedWork W3126277921 @default.
- W3140941716 hasRelatedWork W3127732896 @default.
- W3140941716 hasRelatedWork W396164270 @default.
- W3140941716 hasRelatedWork W4210446059 @default.
- W3140941716 hasRelatedWork W4231964008 @default.
- W3140941716 isParatext "false" @default.
- W3140941716 isRetracted "false" @default.
- W3140941716 magId "3140941716" @default.
- W3140941716 workType "article" @default.