DC Field | Value | Language |
---|---|---|
dc.contributor.author | Park, Jongmin | - |
dc.contributor.author | LEE, YOUNGJOO | - |
dc.date.accessioned | 2021-06-01T06:09:49Z | - |
dc.date.available | 2021-06-01T06:09:49Z | - |
dc.date.created | 2021-03-07 | - |
dc.date.issued | 2020-10-23 | - |
dc.identifier.uri | https://oasis.postech.ac.kr/handle/2014.oak/105839 | - |
dc.description.abstract | Targeting the on-device speech-To-Text application for streaming inputs, this paper presents an efficient way to reduce the computational complexity of deep neural networks (DNNs) for attention-based speech processing. The proposed technique applies the singular value decomposition (SVD) to the large-sized matrix multiplications, removing less important computations by utilizing the low-rank approximation. The clipping thresholds are carefully adjusted to relax the computing costs as well as the memory overheads while maintaining the recognition accuracy. | - |
dc.language | English | - |
dc.publisher | Institute of Electrical and Electronics Engineers Inc. | - |
dc.relation.isPartOf | 17th International System-on-Chip Design Conference, ISOCC 2020 | - |
dc.relation.isPartOf | Proceedings - International SoC Design Conference, ISOCC 2020 | - |
dc.title | Low-Complexity DNN-Based End-To-End Automatic Speech Recognition using Low-Rank Approximation | - |
dc.type | Conference | - |
dc.type.rims | CONF | - |
dc.identifier.bibliographicCitation | 17th International System-on-Chip Design Conference, ISOCC 2020, pp.210 - 211 | - |
dc.identifier.wosid | 000680824100102 | - |
dc.citation.conferenceDate | 2020-10-21 | - |
dc.citation.conferencePlace | KO | - |
dc.citation.endPage | 211 | - |
dc.citation.startPage | 210 | - |
dc.citation.title | 17th International System-on-Chip Design Conference, ISOCC 2020 | - |
dc.contributor.affiliatedAuthor | Park, Jongmin | - |
dc.contributor.affiliatedAuthor | LEE, YOUNGJOO | - |
dc.identifier.scopusid | 2-s2.0-85100711274 | - |
dc.description.journalClass | 1 | - |
dc.description.journalClass | 1 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
library@postech.ac.kr Tel: 054-279-2548
Copyrights © by 2017 Pohang University of Science ad Technology All right reserved.