DC Field | Value | Language |
---|---|---|
dc.contributor.author | Kim, Sangyeob | ko |
dc.contributor.author | Yoo, Hoi-Jun | ko |
dc.contributor.author | LEE, JUHYOUNG | ko |
dc.contributor.author | Kang, Sanghoon | ko |
dc.contributor.author | Lee, Jinmook | ko |
dc.date.accessioned | 2020-12-15T22:50:21Z | - |
dc.date.available | 2020-12-15T22:50:21Z | - |
dc.date.created | 2020-12-01 | - |
dc.date.created | 2020-12-01 | - |
dc.date.created | 2020-12-01 | - |
dc.date.issued | 2020-06-16 | - |
dc.identifier.citation | IEEE Symposium on VLSI Circuits, VLSI Circuits 2020 | - |
dc.identifier.uri | http://hdl.handle.net/10203/278516 | - |
dc.description.abstract | An energy efficient Deep-Neural-Network (DNN) learning processor is proposed for on-chip learning and iterative weight pruning (WP). This work has three key features: 1) stochastic coarse-fine pruning reduced computation workload by 99.7% compared with previous WP algorithm while maintaining high weight sparsity, 2) adaptive input/output/weight skipping (AIOWS) achieved 30.1× higher throughput than previous DNN learning processor [1] for not only the inference but also learning, 3) weight memory shared pruning unit removed on-chip weight memory access for WP. As a result, this work shows 146.52 TOPS/W energy efficiency, which is 5.79× higher than the state-of-the-art [1]. | - |
dc.language | English | - |
dc.publisher | Institute of Electrical and Electronics Engineers Inc. | - |
dc.title | A 146.52 TOPS/W Deep-Neural-Network Learning Processor with Stochastic Coarse-Fine Pruning and Adaptive Input/Output/Weight Skipping | - |
dc.type | Conference | - |
dc.identifier.wosid | 000621657500022 | - |
dc.identifier.scopusid | 2-s2.0-85090236046 | - |
dc.type.rims | CONF | - |
dc.citation.publicationname | IEEE Symposium on VLSI Circuits, VLSI Circuits 2020 | - |
dc.identifier.conferencecountry | US | - |
dc.identifier.conferencelocation | Honolulu, HI | - |
dc.identifier.doi | 10.1109/VLSICircuits18222.2020.9162795 | - |
dc.contributor.localauthor | Yoo, Hoi-Jun | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.