DC Field | Value | Language |
---|---|---|
dc.contributor.author | Choi, Seungkyu | ko |
dc.contributor.author | Sim, Jaehyeong | ko |
dc.contributor.author | Kang, Myeonggu | ko |
dc.contributor.author | Choi, Yeongjae | ko |
dc.contributor.author | Kim, Hyeonuk | ko |
dc.contributor.author | Kim, Lee-Sup | ko |
dc.date.accessioned | 2019-12-13T07:38:12Z | - |
dc.date.available | 2019-12-13T07:38:12Z | - |
dc.date.created | 2019-08-19 | - |
dc.date.created | 2019-08-19 | - |
dc.date.created | 2019-08-19 | - |
dc.date.created | 2019-08-19 | - |
dc.date.issued | 2019-11-05 | - |
dc.identifier.citation | 2019 IEEE Asian Solid-State Circuits Conference | - |
dc.identifier.uri | http://hdl.handle.net/10203/269001 | - |
dc.description.abstract | A scalable deep learning accelerator supporting both inference and training is implemented for device personalization of deep convolutional neural networks. It consists of three processor cores operating with distinct energy-efficient dataflow for different types of computation in CNN training. Two cores conduct forward and backward propagation in convolutional layers and utilize a masking scheme to reduce 88.3% of intermediate data to store for training. The third core executes weight update process in convolutional layers and inner product computation in fully connected layers with a novel large window dataflow. The system enables 8-bit fixed point datapath with lossless training and consumes 47.4J/epoch for a customized deep CNN model. | - |
dc.language | English | - |
dc.publisher | IEEE/SSCS | - |
dc.title | A 47.4µJ/epoch Trainable Deep Convolutional Neural Network Accelerator for In-Situ Personalization on Smart Devices | - |
dc.type | Conference | - |
dc.identifier.wosid | 000569524500017 | - |
dc.identifier.scopusid | 2-s2.0-85108985295 | - |
dc.type.rims | CONF | - |
dc.citation.publicationname | 2019 IEEE Asian Solid-State Circuits Conference | - |
dc.identifier.conferencecountry | CA | - |
dc.identifier.conferencelocation | The Parisian Macao, Cotai Central | - |
dc.contributor.localauthor | Kim, Lee-Sup | - |
dc.contributor.nonIdAuthor | Kang, Myeonggu | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.