• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to previous page

Neural network LSI for deep learning

Research Project

Project/Area Number 17K00083
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Research Field Computer system
Research InstitutionKumamoto University

Principal Investigator

Amagasaki Motoki  熊本大学, 大学院先端科学研究部(工), 准教授 (50467974)

Project Period (FY) 2017-04-01 – 2020-03-31
Project Status Completed (Fiscal Year 2019)
Budget Amount *help
¥4,420,000 (Direct Cost: ¥3,400,000、Indirect Cost: ¥1,020,000)
Fiscal Year 2019: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2018: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
Fiscal Year 2017: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
KeywordsAIチップ / ニューラルネットワーク / DNNアクセラレータ / エッジ端末 / リコンフィギャラブル / 深層学習 / 重み2のべき乗化 / 高速シリアル通信 / ディープラーニング / リコンフィギャラブルシステム / ニューラルネットワークチップ
Outline of Final Research Achievements

In this research, in order to deal with a wide variety of applications and structures of DNNs, we have researched and developed an AI architecture with low cost, high usability, high speed, and ultra-low cost with reconfigurability. The circuit structure of a power-consuming reconfigurable AI accelerator is revealed. We also evaluated the performance at the layout design level using the standard cell library. Maximum operating frequency when implementing the pythorch-AlexNet model for CIFAR100 The frequency was 350 MHz. The processing power per second of the inference model is 100 [FPS], the power consumption is 0.11 [W], and estimated energy efficiency was 883 [FPS/W]. Based on the architecture developed in this research, we plan to work on models for edge computing.

Academic Significance and Societal Importance of the Research Achievements

提案する深層学習チップの特徴は,NNの複雑さや規模に応じてHW構成を最適化できる点にある.特に,演算精度に着目してHW量を削減する回路最適化できる点が重要なポイントである.現在の深層学習は人工知能の先駆けに過ぎず,人間の知能に近づけるには膨大な計算量をいかに高速,低電力でできるかがカギとなる.一方,GPGPUや商用FPGAなどの汎用デバイスを用いたアプローチではこれらに対し限界が来るのは明らかである.提案する深層学習チップ開発を通して,IoTにおけるエッジサイドでの利用に対し、用途に合わせた最適化な形で処理を実行できる枠組みを示した.

Report

(4 results)
  • 2019 Annual Research Report   Final Research Report ( PDF )
  • 2018 Research-status Report
  • 2017 Research-status Report
  • Research Products

    (16 results)

All 2020 2019 2018 2017 Other

All Journal Article (2 results) (of which Peer Reviewed: 2 results,  Open Access: 2 results) Presentation (10 results) Remarks (3 results) Patent(Industrial Property Rights) (1 results)

  • [Journal Article] EVALUATION OF ROUTING AREA REDUCTION FOR FINE-GRAINED OVERLAY VIRTUAL FPGA2020

    • Author(s)
      Theingi Myint, Ito Takanori, Motoki Amagasaki, Qian Zhao, Masahiro Iida
    • Journal Title

      International Journal of Innovative Computing, Information and Control

      Volume: 16 Pages: 0-0

    • Related Report
      2019 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] A SLM-based Overlay Architecture for Fine-grained Virtual FPGA2020

    • Author(s)
      Theingi Myint, Motoki Amagasaki, Qian Zhao, Masahiro Iida
    • Journal Title

      IEICE Electronics Express (ELEX)

      Volume: 16 Pages: 0-0

    • NAID

      130007772826

    • Related Report
      2019 Annual Research Report
    • Peer Reviewed / Open Access
  • [Presentation] 演算精度を考慮したDNN設計フレームワークの開発2019

    • Author(s)
      木山真人
    • Organizer
      電子情報通信学会
    • Related Report
      2019 Annual Research Report
  • [Presentation] Performance Evaluation of Fine-grained Virtual FPGA Based on SLM Architecture2019

    • Author(s)
      Theingi. Mint
    • Organizer
      Proc. of IEEE Symposium on COOL Chips 22
    • Related Report
      2019 Annual Research Report
  • [Presentation] A Novel SLM-based Virtual FPGA Overlay Architecture2019

    • Author(s)
      Theingi. Mint
    • Organizer
      Proc. of IEEE International Symposium on Embedded Multicore/Many-core Systems-on-Chip
    • Related Report
      2019 Annual Research Report
  • [Presentation] Characteristic Similarity Using Classical CNN Model2019

    • Author(s)
      Masato Kiyama
    • Organizer
      Proc. of IEEE International Symposium on Embedded Multicore/Many-core Systems-on-Chip
    • Related Report
      2019 Annual Research Report
  • [Presentation] 高速シリアル光通信を用いたCNN分割実装の検討2018

    • Author(s)
      千竈純太郎・中原康宏・尼崎太樹・久我守弘・飯田全広・末吉敏則
    • Organizer
      電子情報通信学会
    • Related Report
      2018 Research-status Report
  • [Presentation] Resources Utilization of Fine-grained Overlay Architecture2018

    • Author(s)
      Theingi Myint(Kumamoto)・Qian Zhao(Kyutech)・Motoki Amagasaki・Masahiro Iida・Toshinori Sueyoshi
    • Organizer
      電子情報通信学会
    • Related Report
      2018 Research-status Report
  • [Presentation] CNN accerarator using power-of-two weight and pruning2018

    • Author(s)
      中原康宏・千竈純太郎・尼崎太樹・飯田全広・久我守弘・末吉敏則
    • Organizer
      電気・情報関係学会九州支部連合大会
    • Related Report
      2018 Research-status Report
  • [Presentation] CNN implementation using High Speed Optical Serial Links2018

    • Author(s)
      千竈純太郎・中原康宏・尼崎太樹・飯田全広・久我守弘・末吉敏則
    • Organizer
      電気・情報関係学会九州支部連合大会
    • Related Report
      2018 Research-status Report
  • [Presentation] 2のべき乗近似とプルーニングを用いたCNN向けFPGAアクセラレータ2018

    • Author(s)
      宇都宮誉博,尼崎太樹,飯田全広,久我守弘,末吉敏則
    • Organizer
      電子情報通信学会
    • Related Report
      2017 Research-status Report
  • [Presentation] 重みの2のべき乗近似を用いたCNNのFPGA実装に関する一検討2017

    • Author(s)
      宇都宮誉博,尼崎太樹,飯田全広,久我守弘,末吉敏則
    • Organizer
      電子情報通信学会
    • Related Report
      2017 Research-status Report
  • [Remarks]

    • URL

      http://www.arch.cs.kumamoto-u.ac.jp/index.html

    • Related Report
      2019 Annual Research Report
  • [Remarks] Arch研究ホームページ

    • Related Report
      2018 Research-status Report
  • [Remarks] 熊本大学コンピュータアーキテクチャ研究室

    • Related Report
      2017 Research-status Report
  • [Patent(Industrial Property Rights)] ニューラルネットワークの回路及びニューラルネットワーク演算方法2019

    • Inventor(s)
      尼崎太樹; 飯田全広; 中原康宏; 千竈純太郎
    • Industrial Property Rights Holder
      尼崎太樹; 飯田全広; 中原康宏; 千竈純太郎
    • Industrial Property Rights Type
      特許
    • Industrial Property Number
      2019-196326
    • Filing Date
      2019
    • Related Report
      2019 Annual Research Report

URL: 

Published: 2017-04-28   Modified: 2021-02-19  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi