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

不完全性を持つ非線形多重スケールデータのためのオンライン解析法の開発と実応用

Research Project

Project/Area Number 18H01446
Research Category

Grant-in-Aid for Scientific Research (B)

Allocation TypeSingle-year Grants
Section一般
Review Section Basic Section 21020:Communication and network engineering-related
Research InstitutionKeio University

Principal Investigator

湯川 正裕  慶應義塾大学, 理工学部(矢上), 准教授 (60462743)

Co-Investigator(Kenkyū-buntansha) 山田 功  東京工業大学, 工学院, 教授 (50230446)
Project Period (FY) 2018-04-01 – 2022-03-31
Project Status Granted (Fiscal Year 2021)
Budget Amount *help
¥15,730,000 (Direct Cost: ¥12,100,000、Indirect Cost: ¥3,630,000)
Fiscal Year 2021: ¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000)
Fiscal Year 2020: ¥4,550,000 (Direct Cost: ¥3,500,000、Indirect Cost: ¥1,050,000)
Fiscal Year 2019: ¥3,770,000 (Direct Cost: ¥2,900,000、Indirect Cost: ¥870,000)
Fiscal Year 2018: ¥3,250,000 (Direct Cost: ¥2,500,000、Indirect Cost: ¥750,000)
Keywordsカーネル法 / オンラインアルゴリズム / 多重スケール / グループスパース / 弱凸関数 / 頑健推定 / Moreau分解 / カーネル適応フィルタ / 再生核 / オンライン学習 / 弱凸正則化 / 凸最適化 / 多重スケール性 / 不完全性 / オンラインデータ解析
Outline of Annual Research Achievements

1. Minimax Concave 損失関数(MC損失関数)に基づく外れ値に頑健なグループスパース信号復元アルゴリズムを提案し、有効性を実証した。MC損失関数は、弱凸関数と線形作用素の合成関数となっており、Minimax Concave正則化を取り扱う従来の数学的枠組みをそのまま用いることができなかった。本研究では、Moreau分解を用いて、主双対分離法で解くことのできる形に再定式化するとともに、収束条件を満たすアルゴリズムパラメータの必要十分条件を明らかにした。特徴抽出問題への応用に関する数値例を行い、既存のロバスト特徴抽出法と比べ、外れ値に対する頑健性が飛躍的に向上することを実証した。本成果は、信号処理分野のトップジャーナルIEEE Transactions on Signal Processing に掲載された。

2. 多カーネル適応フィルタは、複数の再生核を用いることで非線形関数をオンラインで効率的に推定できる。本研究では、再生核のパラメータをフィルタ係数と同時にデータから学習する手法を提案した。提案法の中で、効率的に辞書を構築していくための「マルチスクリーニング法」を新たに考案した。同手法は、スケールの大きな関数から順に辞書に加えていくことで、辞書要素が不必要に増加していくことを抑制できる。その結果、小さい辞書サイズで高精度な推定が実現される。数値例により、LSTMやカーネル適応フィルタの最新の手法に対する優位性を実証した。本成果は、速報性の高いIEEE Accessに掲載された。

Current Status of Research Progress
Current Status of Research Progress

2: Research has progressed on the whole more than it was originally planned.

Reason

(i)外れ値に頑健なグループスパース信号復元アルゴリズムの構築と(ii)再生核パラメータとフィルタ係数の同時更新アルゴリズムの構築が予定どおり完了した。次の段階として、外れ値に加えてガウス性雑音の効果を明示的に考慮した手法の検討まで進んでおり、研究は順調に進んでいる。

Strategy for Future Research Activity

適宜、共同研究者と議論を行いながら進めていく、

Report

(3 results)
  • 2020 Annual Research Report
  • 2019 Annual Research Report
  • 2018 Annual Research Report

Research Products

(26 results)

All 2022 2021 2020 2019 2018 Other

All Int'l Joint Research (4 results) Journal Article (4 results) (of which Int'l Joint Research: 1 results,  Peer Reviewed: 4 results,  Open Access: 1 results) Presentation (14 results) (of which Int'l Joint Research: 9 results) Book (1 results) Remarks (3 results)

  • [Int'l Joint Research] NCU Torun(ポーランド)

    • Related Report
      2020 Annual Research Report
  • [Int'l Joint Research] Nicolaus Copernicus University(ポーランド)

    • Related Report
      2019 Annual Research Report
  • [Int'l Joint Research] Technical University of Berlin/Fraunhofer Heinrich Hertz Institute(ドイツ)

    • Related Report
      2019 Annual Research Report
  • [Int'l Joint Research] Nicolaus Copernicus University(ポーランド)

    • Related Report
      2018 Annual Research Report
  • [Journal Article] Relaxed zero-forcing beamformer under temporally-correlated interference2022

    • Author(s)
      Kono Takehiro、Yukawa Masahiro、Piotrowski Tomasz
    • Journal Title

      Signal Processing

      Volume: 190 Pages: 108323-108323

    • DOI

      10.1016/j.sigpro.2021.108323

    • Related Report
      2020 Annual Research Report
    • Peer Reviewed / Int'l Joint Research
  • [Journal Article] Joint Learning of Model Parameters and Coefficients for Online Nonlinear Estimation2021

    • Author(s)
      Masa-aki Takizawa and Masahiro Yukawa
    • Journal Title

      IEEE Access

      Volume: Vol.9 Pages: 24026-24040

    • DOI

      10.1109/access.2021.3053651

    • Related Report
      2020 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Robust Recovery of Jointly-Sparse Signals Using Minimax Concave Loss Function2021

    • Author(s)
      Suzuki Kyohei、Yukawa Masahiro
    • Journal Title

      IEEE Transactions on Signal Processing

      Volume: 69 Pages: 669-681

    • DOI

      10.1109/tsp.2020.3044445

    • Related Report
      2020 Annual Research Report
    • Peer Reviewed
  • [Journal Article] Projection-Based Regularized Dual Averaging for Stochastic Optimization2019

    • Author(s)
      Ushio Asahi、Yukawa Masahiro
    • Journal Title

      IEEE Transactions on Signal Processing

      Volume: 67 Pages: 2720-2733

    • DOI

      10.1109/tsp.2019.2908901

    • Related Report
      2019 Annual Research Report
    • Peer Reviewed
  • [Presentation] Robust jointly-sparse signal recovery based on minimax concave loss function2021

    • Author(s)
      Kyohei Suzuki and Masahiro Yukawa
    • Organizer
      European Signal Processing Conference
    • Related Report
      2020 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Steepening squared error function facilitates online adaptation of Gaussian scales2020

    • Author(s)
      Masa-aki Takizawa and Masahiro Yukawa
    • Organizer
      IEEE International Conference on Acoustics, Speech, and Signal Processing
    • Related Report
      2020 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Normalized least-mean-square algorithms with minimax concave penalty2020

    • Author(s)
      Hiroyuki Kaneko and and Masahiro Yukawa
    • Organizer
      IEEE International Conference on Acoustics, Speech, and Signal Processing
    • Related Report
      2020 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Sentiment analysis based on multikernel adaptive filtering: an online approach to multiscale data classification2019

    • Author(s)
      Ran Iwamoto, Masahiro Yukawa
    • Organizer
      European Signal Processing Conference
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Online learning with self-tuned Gaussian kernels: good kernel-initialization by multiscale screening2019

    • Author(s)
      Masaaki Takizawa and Masahiro Yukawa
    • Organizer
      IEEE International Conference on Acoustic, Speech, and Signal Processing
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Automatic kernel weighting for multikernel adaptive filtering: multiscale aspects2019

    • Author(s)
      Kwangjin Jeong and Masahiro Yukawa
    • Organizer
      IEEE International Conference on Acoustic, Speech, and Signal Processing
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Beamformer design under time-correlated interference and online implementation: brain-activity reconstruction from EEG2019

    • Author(s)
      Takehiro Kono, Masahiro Yukawa, and Tomasz Piotrowski
    • Organizer
      IEEE International Conference on Acoustic, Speech, and Signal Processing
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Auto-antonymの多カーネルオンライン意味変化分析2019

    • Author(s)
      岩本蘭、湯川正裕
    • Organizer
      言語処理学会第25回年次大会
    • Related Report
      2018 Annual Research Report
  • [Presentation] An Efficient Online Learning Method Based on Self-tuned Gaussian Kernels2019

    • Author(s)
      Masa-aki Takizawa and Masahiro Yukawa
    • Organizer
      Technical Report of IEICE
    • Related Report
      2018 Annual Research Report
  • [Presentation] Beamforming for Brain-Activity Reconstruction under Time-Correlated Interference2019

    • Author(s)
      Takehiro Kono, Masahiro Yukawa, Tomasz Piotrowski
    • Organizer
      Technical Report of IEICE
    • Related Report
      2018 Annual Research Report
  • [Presentation] Continuous-time Value Function Approximation in Reproducing Kernel Hilbert Spaces2018

    • Author(s)
      Motoya Ohnishi, Masahiro Yukawa, Mikael Johansson, Masashi Sugiyama
    • Organizer
      Advances in Neural Information Processing Systems
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research
  • [Presentation] A Fixed-Point Analysis of Regularized Dual Averaging Under Static Scenarios2018

    • Author(s)
      Masahiro Yukawa and Isao Yamada
    • Organizer
      APSIPA Annual Summit and Conference
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Sentiment Analysis in Twitter Based on Multikernel Adaptive Filtering2018

    • Author(s)
      Ran Iwamoto and Masahiro Yukawa
    • Organizer
      IEICE Signal Processing Symposium
    • Related Report
      2018 Annual Research Report
  • [Presentation] Clustering without Pre-specifying Cluster-Number Based on Sparse Eigenproblem of Graph Laplacian2018

    • Author(s)
      Yuto Ogino and Masahiro Yukawa
    • Organizer
      IEICE Signal Processing Symposium
    • Related Report
      2018 Annual Research Report
  • [Book] Machine Learning for Future Wireless Communications (Chapter 11: Adaptive Learning for Symbol Detection: A Reproducing Kernel Hilbert Space Approach)2020

    • Author(s)
      Daniyal Amir Awan, Renato Luis Garrido Cavalcante, Masahiro Yukawa, and Slawomir Stanczak
    • Publisher
      Wiley
    • ISBN
      9781119562252
    • Related Report
      2019 Annual Research Report
  • [Remarks] Masahiro Yukawa's Page

    • URL

      http://www.ykw.elec.keio.ac.jp/yukawa/

    • Related Report
      2020 Annual Research Report
  • [Remarks] Masahiro Yukawa (Ph.D.)

    • URL

      http://www.ykw.elec.keio.ac.jp/yukawa/

    • Related Report
      2019 Annual Research Report
  • [Remarks] Masahiro Yukawa

    • URL

      http://www.ykw.elec.keio.ac.jp/yukawa/index.html

    • Related Report
      2018 Annual Research Report

URL: 

Published: 2018-04-23   Modified: 2022-12-28  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi