DC Field | Value | Language |
---|---|---|
dc.contributor.author | Shin, Dongjoo | ko |
dc.contributor.author | Lee, Jinmook | ko |
dc.contributor.author | Lee, Jinsu | ko |
dc.contributor.author | Lee, Juhyoung | ko |
dc.contributor.author | Yoo, Hoi-Jun | ko |
dc.date.accessioned | 2018-11-12T04:20:00Z | - |
dc.date.available | 2018-11-12T04:20:00Z | - |
dc.date.created | 2018-10-22 | - |
dc.date.created | 2018-10-22 | - |
dc.date.issued | 2018-09 | - |
dc.identifier.citation | IEEE MICRO, v.38, no.5, pp.85 - 93 | - |
dc.identifier.issn | 0272-1732 | - |
dc.identifier.uri | http://hdl.handle.net/10203/246342 | - |
dc.description.abstract | An energy-efficient deep-learning processor called DNPU is proposed for the embedded processing of convolutional neural networks (CNNs) and recurrent neural networks (RNNs) in mobile platforms. DNPU uses a heterogeneous multi-core architecture to maximize energy efficiency in both CNNs and RNNs. In each core, a memory architecture, data paths, and processing elements are optimized depending on the characteristics of each network. Also, a mixed workload division method is proposed to minimize off-chip memory access in CNNs, and a quantization table-based matrix multiplier is proposed to remove duplicated multiplications in RNNs. | - |
dc.language | English | - |
dc.publisher | IEEE COMPUTER SOC | - |
dc.title | DNPU: An Energy-Efficient Deep-Learning Processor with Heterogeneous Multi-Core Architecture | - |
dc.type | Article | - |
dc.identifier.wosid | 000446337400012 | - |
dc.identifier.scopusid | 2-s2.0-85054524392 | - |
dc.type.rims | ART | - |
dc.citation.volume | 38 | - |
dc.citation.issue | 5 | - |
dc.citation.beginningpage | 85 | - |
dc.citation.endingpage | 93 | - |
dc.citation.publicationname | IEEE MICRO | - |
dc.identifier.doi | 10.1109/MM.2018.053631145 | - |
dc.contributor.localauthor | Yoo, Hoi-Jun | - |
dc.contributor.nonIdAuthor | Lee, Jinsu | - |
dc.contributor.nonIdAuthor | Lee, Juhyoung | - |
dc.description.isOpenAccess | N | - |
dc.type.journalArticle | Article | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.