DC Field | Value | Language |
---|---|---|
dc.contributor.author | Lee, Jinmook | ko |
dc.contributor.author | Shin, Dongjoo | ko |
dc.contributor.author | Yoo, Hoi-Jun | ko |
dc.date.accessioned | 2019-04-15T14:38:24Z | - |
dc.date.available | 2019-04-15T14:38:24Z | - |
dc.date.created | 2018-12-19 | - |
dc.date.created | 2018-12-19 | - |
dc.date.created | 2018-12-19 | - |
dc.date.issued | 2017-11 | - |
dc.identifier.citation | 13th IEEE Asian Solid-State Circuits Conference (A-SSCC), pp.237 - 240 | - |
dc.identifier.uri | http://hdl.handle.net/10203/254242 | - |
dc.description.abstract | A 21mW low-power embedded Recurrent Neural Network (RNN) accelerator is proposed to realize the image captioning applications. The low-power RNN operation is achieved by 3 key features: 1) Quantization-table-based matrix multiplication with RNN weight quantization, 2) Dynamic quantization-table allocation scheme for balanced pipelined RNN operation, and 3) Zero-skipped RNN operation using quantization-table. The Quantization table enables the 98% reduction of the multiplier operations by replacing the multiplication to the table reference. The dynamic quantization table allocation is used to achieve high chip-utilization efficiency over 90% by balanced pipeline operation for three variations of the RNN operation. The zero-skipped RNN operation reduces the overall 27% of required external memory bandwidth and quantization-table operations without any additional hardware cost. The proposed RNN accelerator of 1.84mm(2) achieves 21mW power consumption and demonstrates its functionality on the image captioning RNN in 65nm CMOS process. | - |
dc.language | English | - |
dc.publisher | IEEE Asian Solid-State Circuits Conference 2017 | - |
dc.title | A 21mW Low-power Recurrent Neural Network Accelerator with Quantization Tables for Embedded Deep Learning Applications | - |
dc.type | Conference | - |
dc.identifier.wosid | 000426511300060 | - |
dc.identifier.scopusid | 2-s2.0-85045724584 | - |
dc.type.rims | CONF | - |
dc.citation.beginningpage | 237 | - |
dc.citation.endingpage | 240 | - |
dc.citation.publicationname | 13th IEEE Asian Solid-State Circuits Conference (A-SSCC) | - |
dc.identifier.conferencecountry | KO | - |
dc.identifier.conferencelocation | Grand Hilton Hotel, Seoul | - |
dc.identifier.doi | 10.1109/ASSCC.2017.8240260 | - |
dc.contributor.localauthor | Yoo, Hoi-Jun | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.