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

Representation learning through graph embedding of multi-domain relational data

Research Project

Project/Area Number 20H04148
Research Category

Grant-in-Aid for Scientific Research (B)

Allocation TypeSingle-year Grants
Section一般
Review Section Basic Section 60030:Statistical science-related
Research InstitutionKyoto University

Principal Investigator

SHIMODAIRA Hidetoshi  京都大学, 情報学研究科, 教授 (00290867)

Project Period (FY) 2020-04-01 – 2023-03-31
Project Status Completed (Fiscal Year 2022)
Budget Amount *help
¥17,810,000 (Direct Cost: ¥13,700,000、Indirect Cost: ¥4,110,000)
Fiscal Year 2022: ¥5,460,000 (Direct Cost: ¥4,200,000、Indirect Cost: ¥1,260,000)
Fiscal Year 2021: ¥5,460,000 (Direct Cost: ¥4,200,000、Indirect Cost: ¥1,260,000)
Fiscal Year 2020: ¥6,890,000 (Direct Cost: ¥5,300,000、Indirect Cost: ¥1,590,000)
Keywords多変量解析 / 次元削減 / 分散表現 / 表現学習 / グラフ埋め込み / 自然言語処理 / ニューラルネットワーク / 加法構成性 / パターン認識
Outline of Research at the Start

画像,タグ,文書などの多様なドメインのマルチモーダルデータから関連性を考慮してニューラルネットによって情報統合し共通空間で表現する新しい柔軟な多変量解析の方法論を提案・発展させる.画像認識や自然言語処理などの大規模データに取り組み,その経験を理論にフィードバックすることによって,次元削減したベクトルの加減算など共通空間の演算による推論の枠組みの構築を目指し,高度な思考を実現するステップとする.

Outline of Final Research Achievements

We conducted research to understand how embeddings of relational data represent information. Specifically, We investigated the additive compositionality that forms the basis of analogy calculations using embedded vectors and examined the properties of the embeddings related to it. In conventional additive compositionality, the sum of vectors represents the simultaneous existence of both meanings (AND). However, I demonstrated that the existence of either meaning (OR) can be represented by a frequency-weighted centroid, and the negation of meaning (NOT) is indicated by the negative direction when the origin is relocated to the centroid of the target word set. Furthermore, We theoretically and experimentally demonstrated that the squared norm of word vectors obtained through a form of contrastive learning (SGNS) can be approximated by the Kullback-Leibler (KL) divergence and represents the "strength of meaning."

Academic Significance and Societal Importance of the Research Achievements

本研究は,関連性データの埋め込みと表現学習に関する新たな知見を提供しました.加法構成性や埋め込みの性質に関する結果は,単語や概念をベクトルで表現する方法に関する理論的な理解を深めることに貢献しました.これにより,なぜニューラルネットが効果的に機能するのか,その原理を理解する道を開くことが期待されます.

Report

(4 results)
  • 2022 Annual Research Report   Final Research Report ( PDF )
  • 2021 Annual Research Report
  • 2020 Annual Research Report
  • Research Products

    (17 results)

All 2022 2021 2020

All Journal Article (4 results) (of which Int'l Joint Research: 1 results,  Peer Reviewed: 4 results,  Open Access: 3 results) Presentation (13 results) (of which Int'l Joint Research: 1 results)

  • [Journal Article] Selective inference after feature selection via multiscale bootstrap2022

    • Author(s)
      Terada Yoshikazu、Shimodaira Hidetoshi
    • Journal Title

      Annals of the Institute of Statistical Mathematics

      Volume: 75 Issue: 1 Pages: 99-125

    • DOI

      10.1007/s10463-022-00838-2

    • Related Report
      2022 Annual Research Report
    • Peer Reviewed
  • [Journal Article] A Hypergraph Approach for Estimating Growth Mechanisms of Complex Networks2022

    • Author(s)
      Inoue Masaaki、Pham Thong、Shimodaira Hidetoshi
    • Journal Title

      IEEE Access

      Volume: 10 Pages: 35012-35025

    • DOI

      10.1109/access.2022.3143612

    • Related Report
      2021 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Non-parametric estimation of the preferential attachment function from one network snapshot2021

    • Author(s)
      Pham Thong、Sheridan Paul、Shimodaira Hidetoshi
    • Journal Title

      Journal of Complex Networks

      Volume: 9 Issue: 5

    • DOI

      10.1093/comnet/cnab024

    • Related Report
      2021 Annual Research Report
    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] Extrapolation Towards Imaginary 0-Nearest Neighbour and Its Improved Convergence Rate2020

    • Author(s)
      Akifumi Okuno, Hidetoshi Shimodaira
    • Journal Title

      Advances in Neural Information Processing Systems (NeurIPS 2020)

      Volume: 33 Pages: 21889-21899

    • Related Report
      2020 Annual Research Report
    • Peer Reviewed / Open Access
  • [Presentation] 単語ベクトルの長さはKL情報量で解釈可能な意味の強さを表す2022

    • Author(s)
      大山百々勢, 横井祥, 下平英寿
    • Organizer
      第25回情報論的学習理論ワークショップ (IBIS2022)
    • Related Report
      2022 Annual Research Report
  • [Presentation] Heterogeneous-featureを用いたマルチタスク転移学習2022

    • Author(s)
      Runsen Li, 奥野彰文, 下平英寿
    • Organizer
      第25回情報論的学習理論ワークショップ (IBIS2022)
    • Related Report
      2022 Annual Research Report
  • [Presentation] 単語ベクトルの長さは意味の強さを表す2022

    • Author(s)
      大山百々勢,横井祥,下平英寿
    • Organizer
      言語処理学会第28回年次大会(NLP2022)
    • Related Report
      2021 Annual Research Report
  • [Presentation] Revisiting Additive Compositionality: AND, OR and NOT Operations with Word Embeddings2021

    • Author(s)
      Masahiro Naito , Sho Yokoi , Geewook Kim , Hidetoshi Shimodaira
    • Organizer
      ACL-IJCNLP 2021 Student Research Workshop
    • Related Report
      2021 Annual Research Report
    • Int'l Joint Research
  • [Presentation] マルチスケールk-近傍法における回帰関数および損失関数の検討2021

    • Author(s)
      CAO Ruixing , 田中卓磨 , 奥野彰文 , 下平英寿
    • Organizer
      第35回人工知能学会全国大会 (JSAI2021)
    • Related Report
      2021 Annual Research Report
  • [Presentation] マルチスケールk-近傍法による画像のExtreme Multi-Label分類2021

    • Author(s)
      田中卓磨 , 奥野彰文 , 下平英寿
    • Organizer
      第35回人工知能学会全国大会 (JSAI2021)
    • Related Report
      2021 Annual Research Report
  • [Presentation] マルチスケールk-近傍法における外挿モデルの検討2021

    • Author(s)
      操瑞行 , 田中卓磨 , 奥野彰文 , 下平英寿
    • Organizer
      2021年度統計関連学会連合大会
    • Related Report
      2021 Annual Research Report
  • [Presentation] 単語埋め込みの加法構成性の精緻化と論理演算2021

    • Author(s)
      内藤雅博 , 横井祥 , 下平英寿
    • Organizer
      2021年度統計関連学会連合大会
    • Related Report
      2021 Annual Research Report
  • [Presentation] 任意のノード特徴量による成長機構をもつハイパーグラフモデル2021

    • Author(s)
      井上雅章 , THONG Pham , 下平英寿
    • Organizer
      2021年度統計関連学会連合大会
    • Related Report
      2021 Annual Research Report
  • [Presentation] 複雑ネットワークの成長過程を観測できない時の優先的選択関数の推定方法2021

    • Author(s)
      THONG Pham , PAUL Sheridan , 下平英寿
    • Organizer
      2021年度統計関連学会連合大会
    • Related Report
      2021 Annual Research Report
  • [Presentation] 単語埋め込みによる論理演算2021

    • Author(s)
      内藤雅博, 横井祥, 下平英寿
    • Organizer
      言語処理学会第27回年次大会(NLP2021)
    • Related Report
      2020 Annual Research Report
  • [Presentation] 単語埋め込みの確率的等方化2021

    • Author(s)
      横井祥, 下平英寿
    • Organizer
      言語処理学会第27回年次大会(NLP2021)
    • Related Report
      2020 Annual Research Report
  • [Presentation] 仮想的なゼロ近傍への外挿とその収束レートについて2020

    • Author(s)
      奥野 彰文,下平 英寿
    • Organizer
      2020年度統計関連学会連合
    • Related Report
      2020 Annual Research Report

URL: 

Published: 2020-04-28   Modified: 2024-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi