NN compactor: Minimizing memory and logic resources for small neural networks
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Hong, Seongmin | - |
dc.contributor.author | Lee, Inho | - |
dc.contributor.author | Park, Yongjun | - |
dc.date.accessioned | 2022-07-12T00:47:51Z | - |
dc.date.available | 2022-07-12T00:47:51Z | - |
dc.date.created | 2021-05-13 | - |
dc.date.issued | 2018-04 | - |
dc.identifier.uri | https://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/150229 | - |
dc.description.abstract | Special neural accelerators are an appealing hardware platform for machine learning systems because they provide both high performance and energy efficiency. Although various neural accelerators have recently been introduced, they are difficult to adapt to embedded platforms because current neural accelerators require high memory capacity and bandwidth for the fast preparation of synaptic weights. Embedded platforms are often unable to meet these memory requirements because of their limited resources. In FPGA-based IoT (internet of things) systems, the problem becomes even worse since computation units generated from logic blocks cannot be fully utilized due to the small size of block memory. In order to overcome this problem, we propose a novel dual-track quantization technique to reduce synaptic weight width based on the magnitude of the value while minimizing accuracy loss. In this value-adaptive technique, large and small value weights are quantized differently. In this paper, we present a fully automatic framework called NN Compactor that generates a compact neural accelerator by minimizing the memory requirements of synaptic weights through dual-track quantization and minimizing the logic requirements of PUs with minimum recognition accuracy loss. For the three widely used datasets of MNIST, CNAE-9, and Forest, experimental results demonstrate that our compact neural accelerator achieves an average performance improvement of 6.4χ over a baseline embedded system using minimal resources with minimal accuracy loss. | - |
dc.language | 영어 | - |
dc.language.iso | en | - |
dc.publisher | Institute of Electrical and Electronics Engineers Inc. | - |
dc.title | NN compactor: Minimizing memory and logic resources for small neural networks | - |
dc.type | Article | - |
dc.contributor.affiliatedAuthor | Park, Yongjun | - |
dc.identifier.doi | 10.23919/DATE.2018.8342074 | - |
dc.identifier.scopusid | 2-s2.0-85048760007 | - |
dc.identifier.bibliographicCitation | Proceedings of the 2018 Design, Automation and Test in Europe Conference and Exhibition, DATE 2018, v.2018-January, pp.581 - 584 | - |
dc.relation.isPartOf | Proceedings of the 2018 Design, Automation and Test in Europe Conference and Exhibition, DATE 2018 | - |
dc.citation.title | Proceedings of the 2018 Design, Automation and Test in Europe Conference and Exhibition, DATE 2018 | - |
dc.citation.volume | 2018-January | - |
dc.citation.startPage | 581 | - |
dc.citation.endPage | 584 | - |
dc.type.rims | ART | - |
dc.type.docType | Conference Paper | - |
dc.description.journalClass | 1 | - |
dc.description.isOpenAccess | N | - |
dc.description.journalRegisteredClass | scopus | - |
dc.subject.keywordPlus | Acceleration | - |
dc.subject.keywordPlus | Automation | - |
dc.subject.keywordPlus | Computation theory | - |
dc.subject.keywordPlus | Energy efficiency | - |
dc.subject.keywordPlus | Internet of things | - |
dc.subject.keywordPlus | Learning systems | - |
dc.subject.keywordPlus | Neural networks | - |
dc.subject.keywordPlus | Particle accelerators | - |
dc.subject.keywordPlus | Adaptive technique | - |
dc.subject.keywordPlus | Embedded platforms | - |
dc.subject.keywordPlus | Hardware platform | - |
dc.subject.keywordPlus | Logic resources | - |
dc.subject.keywordPlus | Memory requirements | - |
dc.subject.keywordPlus | Performance improvements | - |
dc.subject.keywordPlus | Quantization | - |
dc.subject.keywordPlus | Recognition accuracy | - |
dc.subject.keywordPlus | Computer circuits | - |
dc.subject.keywordAuthor | Accelerator | - |
dc.subject.keywordAuthor | Automation | - |
dc.subject.keywordAuthor | Neural networks | - |
dc.subject.keywordAuthor | Quantization | - |
dc.identifier.url | https://ieeexplore.ieee.org/document/8342074 | - |
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.
222, Wangsimni-ro, Seongdong-gu, Seoul, 04763, Korea+82-2-2220-1365
COPYRIGHT © 2021 HANYANG UNIVERSITY.
Certain data included herein are derived from the © Web of Science of Clarivate Analytics. All rights reserved.
You may not copy or re-distribute this material in whole or in part without the prior written consent of Clarivate Analytics.