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

Hypercomplex-valued Deep Neural Networks and Their Applications to Image Analysis

Research Project

Project/Area Number 16K00248
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Research Field Perceptual information processing
Research InstitutionUniversity of Hyogo

Principal Investigator

Isokawa Teijiro  兵庫県立大学, 工学研究科, 准教授 (70336832)

Co-Investigator(Kenkyū-buntansha) 松井 伸之  兵庫県立大学, 工学研究科, 特任教授 (10173783)
Project Period (FY) 2016-04-01 – 2019-03-31
Project Status Completed (Fiscal Year 2018)
Budget Amount *help
¥4,420,000 (Direct Cost: ¥3,400,000、Indirect Cost: ¥1,020,000)
Fiscal Year 2018: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2017: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2016: ¥2,210,000 (Direct Cost: ¥1,700,000、Indirect Cost: ¥510,000)
Keywordsニューラルネットワーク / 複素ニューラルネットワーク / 四元数 / エクストリーム学習器 / 畳み込みニューラルネットワークモデル / 深層学習 / 可換四元数
Outline of Final Research Achievements

The purpose of this project is to construct deep neural networks based on hypercomplex-value systems and to evaluate the performances of these neural networks through the tasks of image analysis. It is expected that the representations and operations in hypercomplex number systems would be beneficial to multidimensional data processing, as compared to conventional (real-valued) neural networks.
The following outcomes have been achieved in this project: (1) Quaternionic (one of hypercomplex-value systems) Extreme Learning Machine (QELM) and quaternionic convolutional neural networks have been proposed. (2) QELM has been implemented on a reconfigurable processor and its performance is evaluated through the problem of color information retrieval from the images under low illuminations.

Academic Significance and Societal Importance of the Research Achievements

従来のニューラルネットワークを大規模化したものは深層学習あるいはディープラーニングと呼ばれるものであり,これは画像情報などの多次元データを多数のニューロンと呼ばれる基本素子により処理するシステムである.本研究課題では,ニューロンの数ではなく各ニューロンが多次元のデータを処理することにより大規模化する方法を検討したものである.本課題において構成したニューラルネットワークでは,多次元のデータを処理するために多次元の数体系を導入することにより,従来の実数に基づくニューラルネットワークよりも効率的に処理できうることを示し得た.

Report

(4 results)
  • 2018 Annual Research Report   Final Research Report ( PDF )
  • 2017 Research-status Report
  • 2016 Research-status Report
  • Research Products

    (11 results)

All 2019 2018 2017 2016

All Journal Article (3 results) (of which Int'l Joint Research: 1 results,  Peer Reviewed: 3 results,  Open Access: 2 results,  Acknowledgement Compliant: 1 results) Presentation (8 results) (of which Int'l Joint Research: 3 results)

  • [Journal Article] Heterogeneous recurrent neural networks for natural language model2019

    • Author(s)
      M.Tsuji, T.Isokawa, N.Yumoto, N.Matsui, and N.Kamiura
    • Journal Title

      Artificial Life and Robotics

      Volume: 未定 Issue: 2 Pages: 1-5

    • DOI

      10.1007/s10015-018-0507-1

    • Related Report
      2018 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Complex-Valued Associative Memories with Projection and Iterative Learning Rules2018

    • Author(s)
      Isokawa Teijiro, Yamamoto Hiroki, Nishimura Haruhiko, Yumoto Takayuki, Kamiura Naotake, Matsui Nobuyuki
    • Journal Title

      Journal of Artificial Intelligence and Soft Computing Research

      Volume: 8 Issue: 3 Pages: 237-249

    • DOI

      10.1515/jaiscr-2018-0015

    • Related Report
      2017 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Feed forward neural network with random quaternionic neurons2017

    • Author(s)
      T.Minemoto, T.Isokawa, H.Nishimura, and N.Matsui
    • Journal Title

      Signal Processing

      Volume: 136 Pages: 59-68

    • DOI

      10.1016/j.sigpro.2016.11.008

    • Related Report
      2016 Research-status Report
    • Peer Reviewed / Int'l Joint Research / Acknowledgement Compliant
  • [Presentation] 量子化QELMニューラルネットワークによるカラーナイトビジョンシステムとIoT向けDSPにおける実装・評価2018

    • Author(s)
      藤井航基, 礒川悌次郎
    • Organizer
      Cadance User Conference (CDNLive Japan 2018)
    • Related Report
      2018 Annual Research Report
  • [Presentation] 四元数化したエクストリーム学習器のDSP実装とその評価2018

    • Author(s)
      礒川悌次郎
    • Organizer
      計測自動制御学会システム・情報部門学術講演会
    • Related Report
      2018 Annual Research Report
  • [Presentation] A Neural Language Model by Heterogeneous Recurrent Neural Networks2018

    • Author(s)
      Masayuki Tsuji (M.Tsuji, T.Isokawa, T.Yumoto, N.Matsui, and N.Kamiura)
    • Organizer
      23rd International Symposium on Artificial Life and Robotics 2017 (AROB 23rd 2018)
    • Related Report
      2017 Research-status Report
    • Int'l Joint Research
  • [Presentation] 四元数Extreme Learning MachineニューラルネットワークのXtensaへの実装と性能評価2017

    • Author(s)
      藤井航基,礒川悌次郎 (藤井航基, 礒川悌次郎, 松井伸之)
    • Organizer
      Cadance User Conference (CDNLive Japan 2017)
    • Related Report
      2017 Research-status Report
  • [Presentation] QELMニューラルネットワークの性能評価2017

    • Author(s)
      礒川悌次郎 (藤井航基, 峯本俊文, 礒川悌次郎, 松井伸之)
    • Organizer
      電子情報通信学会 第63回機能集積情報システム研究会
    • Related Report
      2017 Research-status Report
  • [Presentation] 機械学習を用いた健康診断における項目値の予測2017

    • Author(s)
      成田健 (成田健, 礒川悌次郎, 松井伸之, 湯本高行, 上浦尚武, 岡本稔, 高山哲郎)
    • Organizer
      計測自動制御学会システム・情報部門学術講演会論文集
    • Related Report
      2017 Research-status Report
  • [Presentation] Pattern Retrieval by Quaternionic Associative Memory with Dual Connections2016

    • Author(s)
      T.Minemoto, T.Isokawa, M.Kobayashi, H.Nishimura, and N.Matsui
    • Organizer
      The 23rd International Conference on Neural Information Processing (ICONIP 2016)
    • Place of Presentation
      京都大学(京都府京都市)
    • Year and Date
      2016-10-16
    • Related Report
      2016 Research-status Report
    • Int'l Joint Research
  • [Presentation] Retrieval Performance of Hopfield Associative Memory with Complex-Valued and Real-Valued Neurons2016

    • Author(s)
      T.Minemoto, T.Isokawa, M.Kobayashi, H.Nishimura, and N.Matsui
    • Organizer
      2016 International Join Conference on Neural Networks (IJCNN2016)
    • Place of Presentation
      バンクーバー(カナダ)
    • Year and Date
      2016-06-24
    • Related Report
      2016 Research-status Report
    • Int'l Joint Research

URL: 

Published: 2016-04-21   Modified: 2020-03-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi