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

A universal system for constructive data preprocessing

Research Project

Project/Area Number 21K11778
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Review Section Basic Section 60030:Statistical science-related
Research InstitutionYamagata University

Principal Investigator

YASUDA Muneki  山形大学, 大学院理工学研究科, 教授 (20532774)

Project Period (FY) 2021-04-01 – 2024-03-31
Project Status Completed (Fiscal Year 2023)
Budget Amount *help
¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000)
Fiscal Year 2023: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2022: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Fiscal Year 2021: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Keywordsデータ前処理 / 特徴抽出 / 確率的ニューラルネットワーク / 統計的機械学習 / スパースモデリング / 統計的近似アルゴリズム / 確率モデル / データ要素重要度分析 / 統計的深層学習 / ベイズ統計 / 近似アルゴリズム / 特徴量抽出 / ボルツマンマシン
Outline of Research at the Start

適切なデータ前処理(または特徴量抽出)はその後のシステム(例えば深層学習システム)の性能に大きな影響を与えることが知られている。しかしながら、ノイズ除去や不要次元の剪定などを含むような「積極的前処理」は課題個別に設計されることが主であり、汎用的に利用できるものは残念ながら存在しない。
本研究は、確率的グラフィカルモデルと統計的機械学習理論を基礎として、汎用的に利用できる積極的データ前処理器の確立を目指す。

Outline of Final Research Achievements

Good data pre-processings are important for various subsequent data science tasks. Therefore, a universal algorithm for data pre-processing is required. The main goal of this research is to build versatile algorithms for constructive (or active) data pre-processings which involve noise reduction and pruning of unwanted dimensions in data. The main results obtained within the research period are as follows.
(1) Fundamental models based on probabilistic neural networks and algorithms handling them, for constructive data pre-processings were constructed.
(2) Through constructing high-quality statistical approximation algorithms and proposal of extension model fused with sparse modelling, we have extensionally developed probabilistic neural networks. They are expected to realize truly general-purpose constructive data pre-processings.

Academic Significance and Societal Importance of the Research Achievements

データ前処理は種々のデータサイエンス課題の成功に対する鍵となるが、良質なデータ前処理の実現には、しばしば分野の専門知識や、それを超えた特別なアイディアが必要となってしまう。最適なデータ前処理アルゴリズムは個々のデータの性質に大きく依存するため、常に最適化な結果を与える万能なデータ前処理アルゴリズムは存在しない。しかしながら、ある程度汎用的に利用できるデータ前処理器ならおそらく実現可能である。特に、ノイズ除去や不要次元の剪定などを含むような積極的前処理はより重要である。積極的前処理を含むような汎用的データ前処理アルゴリズムの存在は、「誰でも成果を出すことができる」の実現を近づけることとなる。

Report

(4 results)
  • 2023 Annual Research Report   Final Research Report ( PDF )
  • 2022 Research-status Report
  • 2021 Research-status Report
  • Research Products

    (47 results)

All 2024 2023 2022 2021 Other

All Journal Article (15 results) (of which Int'l Joint Research: 2 results,  Peer Reviewed: 15 results,  Open Access: 12 results) Presentation (29 results) (of which Int'l Joint Research: 4 results,  Invited: 1 results) Book (2 results) Remarks (1 results)

  • [Journal Article] Effective sampling on Gaussian-Bernoulli restricted Boltzmann machines2024

    • Author(s)
      Muneki Yasuda
    • Journal Title

      Nonlinear Theory and Its Applications, IEICE

      Volume: 15 Issue: 2 Pages: 217-225

    • DOI

      10.1587/nolta.15.217

    • ISSN
      2185-4106
    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Quasi-free energy evaluation of Gaussian-Bernoulli restricted Boltzmann machine for anomaly detection2024

    • Author(s)
      Kaiji Sekimoto, Chako Takahashi, Muneki Yasuda
    • Journal Title

      Nonlinear Theory and Its Applications, IEICE

      Volume: 15 Issue: 2 Pages: 273-283

    • DOI

      10.1587/nolta.15.273

    • ISSN
      2185-4106
    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Gaussian-discrete restricted Boltzmann machine with sparse-regularized hidden layer2024

    • Author(s)
      Kaiji Sekimoto, Muneki Yasuda
    • Journal Title

      Behaviormetrika

      Volume: 51 Issue: 1 Pages: 5-23

    • DOI

      10.1007/s41237-024-00230-9

    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] New Learning Algorithm of Gaussian–Bernoulli Restricted Boltzmann Machine and its Application in Feature Extraction2023

    • Author(s)
      Muneki Yasuda, Zhongren Xiong
    • Journal Title

      IEICE Proceeding Series

      Volume: 76 Pages: 134-137

    • DOI

      10.34385/proc.76.A3L-42

    • ISSN
      2188-5079
    • Year and Date
      2023-09-21
    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Journal Article] Quasi-Free Energy Evaluation of Restricted Boltzmann Machine for Anomaly Detection2023

    • Author(s)
      Kaiji Sekimoto, Chako Takahashi, and Muneki Yasuda
    • Journal Title

      IEICE Proceeding Series

      Volume: 76 Pages: 142-145

    • DOI

      10.34385/proc.76.A3L-44

    • ISSN
      2188-5079
    • Year and Date
      2023-09-21
    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Discriminative restricted Boltzmann machine with trainable sparsity2023

    • Author(s)
      Yasuda Muneki、Katsumata Tomu
    • Journal Title

      Nonlinear Theory and Its Applications, IEICE

      Volume: 14 Issue: 2 Pages: 207-214

    • DOI

      10.1587/nolta.14.207

    • ISSN
      2185-4106
    • Related Report
      2022 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Effective learning algorithm for restricted Boltzmann machines via spatial Monte Carlo integration2023

    • Author(s)
      Sekimoto Kaiji、Yasuda Muneki
    • Journal Title

      Nonlinear Theory and Its Applications, IEICE

      Volume: 14 Issue: 2 Pages: 228-241

    • DOI

      10.1587/nolta.14.228

    • ISSN
      2185-4106
    • Related Report
      2022 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Discriminative Restricted Boltzmann Machine with Adapted-Sparse Hidden Layer2022

    • Author(s)
      Muneki Yasuda、Tomu Katsumata
    • Journal Title

      IEICE Proceeding Series

      Volume: 71 Pages: 53-56

    • DOI

      10.34385/proc.71.A3L-B-01

    • ISSN
      2188-5079
    • Year and Date
      2022-12-12
    • Related Report
      2022 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Spatial Monte Carlo Integration for Learning Restricted Boltzmann Machines2022

    • Author(s)
      Kaiji Sekimoto、Muneki Yasuda
    • Journal Title

      IEICE Proceeding Series

      Volume: 71 Pages: 9-12

    • DOI

      10.34385/proc.71.A2L-B-03

    • ISSN
      2188-5079
    • Year and Date
      2022-12-12
    • Related Report
      2022 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Hierarchical Gaussian Markov Random Field for Image Denoising2022

    • Author(s)
      Yuki Monma, Aro Kan, and Muneki Yasuda
    • Journal Title

      IEICE Transactions on Information and Systems

      Volume: E105.D Issue: 3 Pages: 689-699

    • DOI

      10.1587/transinf.2021EDP7172

    • NAID

      130008165627

    • ISSN
      0916-8532, 1745-1361
    • Year and Date
      2022-03-01
    • Related Report
      2021 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Free energy evaluation using marginalized annealed importance sampling2022

    • Author(s)
      Yasuda Muneki、Takahashi Chako
    • Journal Title

      Physical Review E

      Volume: 106 Issue: 2 Pages: 1-11

    • DOI

      10.1103/physreve.106.024127

    • Related Report
      2022 Research-status Report
    • Peer Reviewed
  • [Journal Article] Composite Spatial Monte Carlo Integration Based on Generalized Least Squares2022

    • Author(s)
      Sekimoto Kaiji、Yasuda Muneki
    • Journal Title

      Journal of the Physical Society of Japan

      Volume: 91 Issue: 11 Pages: 1-12

    • DOI

      10.7566/jpsj.91.114003

    • Related Report
      2022 Research-status Report
    • Peer Reviewed
  • [Journal Article] Spatial Monte Carlo integration with annealed importance sampling2021

    • Author(s)
      Muneki Yasuda and Kaiji Sekimoto
    • Journal Title

      Physical Review E

      Volume: 103 Issue: 5 Pages: 1-10

    • DOI

      10.1103/physreve.103.052118

    • Related Report
      2021 Research-status Report
    • Peer Reviewed
  • [Journal Article] Effective fine-tuning training of deep Boltzmann machine based on spatial Monte Carlo integration2021

    • Author(s)
      Tomu Katsumata and Muneki Yasuda
    • Journal Title

      Nonlinear Theory and Its Applications, IEICE

      Volume: 12 Issue: 3 Pages: 377-390

    • DOI

      10.1587/nolta.12.377

    • NAID

      130008060807

    • ISSN
      2185-4106
    • Related Report
      2021 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] The association between the subjective quality of sleep and the phase of a 90-minute periodic signal for the wake-up support system2021

    • Author(s)
      Seung-Il Cho, Minami Tsuchiya, Atsushi Tanaka, Muneki Yasuda, Tomochika Harada, and Michio Yokoyama
    • Journal Title

      Nonlinear Theory and Its Applications, IEICE

      Volume: 12 Issue: 3 Pages: 464-474

    • DOI

      10.1587/nolta.12.464

    • NAID

      130008060816

    • ISSN
      2185-4106
    • Related Report
      2021 Research-status Report
    • Peer Reviewed / Open Access / Int'l Joint Research
  • [Presentation] 制限ボルツマンマシンとt-SNEを用いた特徴抽出器2024

    • Author(s)
      松川岬矢,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2023 Annual Research Report
  • [Presentation] 深層ボルツマンマシン分類器に対する高性能学習法2024

    • Author(s)
      石沢怜,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2023 Annual Research Report
  • [Presentation] 多層構造学習モデルの中間層に対する情報理論的分析2024

    • Author(s)
      栗林諒,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2023 Annual Research Report
  • [Presentation] 異なるノイズ分散をもつ複数劣化画像からの画像修復2024

    • Author(s)
      佐藤憧,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2023 Annual Research Report
  • [Presentation] KLIEPを用いた相対密度比推定2024

    • Author(s)
      浦田光佑,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2023 Annual Research Report
  • [Presentation] ベイジアンネットワーク型診断システムに対する高効率学習法2024

    • Author(s)
      高橋隼汰,安田宗樹
    • Organizer
      情報処理学会第86回全国大会
    • Related Report
      2023 Annual Research Report
  • [Presentation] V正則化型相互作用をもつマルコフ確率場モデルの提案2024

    • Author(s)
      芳賀友紀,関本快士,安田宗樹
    • Organizer
      情報処理学会第86回全国大会
    • Related Report
      2023 Annual Research Report
  • [Presentation] 階層ベイズ学習に基づく組み合わせ最適化問題の統計的分析2024

    • Author(s)
      石岡龍佑,関本快士,安田宗樹
    • Organizer
      情報処理学会第86回全国大会
    • Related Report
      2023 Annual Research Report
  • [Presentation] ハフ変換とグラフィカルモデルを用いた車線推定2024

    • Author(s)
      渡部直生,酒井佳奈子,土谷千加夫,安田宗樹
    • Organizer
      情報処理学会第86回全国大会
    • Related Report
      2023 Annual Research Report
  • [Presentation] 悪環境下における機械学習2024

    • Author(s)
      安田宗樹
    • Organizer
      2023年度太陽研連シンポジウム
    • Related Report
      2023 Annual Research Report
    • Invited
  • [Presentation] 多次元混合ガウス分布を用いたTwitterユーザ集団の偏り測定2023

    • Author(s)
      高橋茶子,吉田光男,安田宗樹
    • Organizer
      人工知能学会第37回全国大会
    • Related Report
      2023 Annual Research Report
  • [Presentation] New Learning Algorithm of Gaussian-Bernoulli Restricted Boltzmann Machine and its Application in Feature Extraction2023

    • Author(s)
      Muneki Yasuda, Zhongren Xiong
    • Organizer
      The 2023 International Symposium on Nonlinear Theory and its Applications
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Quasi-Free Energy Evaluation of Restricted Boltzmann Machine for Anomaly Detection2023

    • Author(s)
      Kaiji Sekimoto, Chako Takahashi, Muneki Yasuda
    • Organizer
      The 2023 International Symposium on Nonlinear Theory and its Applications
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] 学習済みBERTクラス分類器からのクラス間相関構造の抽出2023

    • Author(s)
      高橋良介,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2022 Research-status Report
  • [Presentation] 制限ボルツマンマシンを用いた特徴量抽出と特徴重要度分析2023

    • Author(s)
      熊中仁,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2022 Research-status Report
  • [Presentation] 制限ボルツマンマシンを用いた欠損のあるデータ集合の学習2023

    • Author(s)
      関本快士,安田宗樹
    • Organizer
      情報処理学会 第85回全国大会
    • Related Report
      2022 Research-status Report
  • [Presentation] 最適化問題における統計的揺らぎの分析2023

    • Author(s)
      石岡龍佑,安田宗樹
    • Organizer
      情報処理学会 第85回全国大会
    • Related Report
      2022 Research-status Report
  • [Presentation] 経験ベイズ法の統計力学的解析の一般化と性能検証2023

    • Author(s)
      高橋隼汰,安田宗樹
    • Organizer
      情報処理学会 第85回全国大会
    • Related Report
      2022 Research-status Report
  • [Presentation] Discriminative Restricted Boltzmann Machine with Adapted-Sparse Hidden Layer2022

    • Author(s)
      Muneki Yasuda、Tomu Katsumata
    • Organizer
      The 2022 International Symposium on Nonlinear Theory and its Applications
    • Related Report
      2022 Research-status Report
    • Int'l Joint Research
  • [Presentation] Spatial Monte Carlo Integration for Learning Restricted Boltzmann Machine2022

    • Author(s)
      Kaiji Sekimoto、Muneki Yasuda
    • Organizer
      The 2022 International Symposium on Nonlinear Theory and its Applications
    • Related Report
      2022 Research-status Report
    • Int'l Joint Research
  • [Presentation] 階層型ガウシアンマルコフ確率場を用いた画像ノイズ除去2022

    • Author(s)
      門馬維紀,安田宗樹
    • Organizer
      情報処理学会 第84回全国大会
    • Related Report
      2021 Research-status Report
  • [Presentation] 損失関数を基礎とした事前分布をもつベイジアンニューラルネットワーク2022

    • Author(s)
      葛原優樹,浦田光佑,安田宗樹
    • Organizer
      情報処理学会 第84回全国大会
    • Related Report
      2021 Research-status Report
  • [Presentation] Cost Sensitive 学習に対する重み付きバッチ正規化と重み付き入力正規化2022

    • Author(s)
      楊顕恩,安田宗樹
    • Organizer
      情報処理学会 第84回全国大会
    • Related Report
      2021 Research-status Report
  • [Presentation] 一般化最小二乗法による合成空間モンテカルロ積分法2022

    • Author(s)
      関本快士,安田宗樹
    • Organizer
      情報処理学会 第84回全国大会
    • Related Report
      2021 Research-status Report
  • [Presentation] マルコフ確率場を用いたパンデミック・リスク予測2022

    • Author(s)
      大高郁斗,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2021 Research-status Report
  • [Presentation] 階層ガウス型マルコフ確率場を用いた画像補修2022

    • Author(s)
      池谷真弥,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2021 Research-status Report
  • [Presentation] 制限ボルツマンマシンを用いたデータ前処理2022

    • Author(s)
      田中佑典,安田宗樹
    • Organizer
      情報処理学会東北支部研究会
    • Related Report
      2021 Research-status Report
  • [Presentation] 一般化最小二乗法による合成空間モンテカルロ積分法2022

    • Author(s)
      関本快士,安田宗樹
    • Organizer
      日本物理学会 第77回年次大会
    • Related Report
      2021 Research-status Report
  • [Presentation] 制限ボルツマンマシンに対する統計力学的解析と学習への応用2022

    • Author(s)
      前野陵介,安田宗樹
    • Organizer
      日本物理学会 第77回年次大会
    • Related Report
      2021 Research-status Report
  • [Book] 機械学習・ディープラーニングによる“異常検知”技術と活用事例集2022

    • Author(s)
      執筆者:70名、技術情報協会
    • Total Pages
      560
    • Publisher
      技術情報協会
    • ISBN
      9784861049132
    • Related Report
      2022 Research-status Report
  • [Book] Sublinear Computation Paradigm2021

    • Author(s)
      Naoki Katoh et. al.
    • Total Pages
      410
    • Publisher
      Springer Singapore
    • ISBN
      9789811640940
    • Related Report
      2021 Research-status Report
  • [Remarks] 成果に関する web ページ

    • URL

      http://www.adv-pip.yz.yamagata-u.ac.jp/~muneki/index.html

    • Related Report
      2022 Research-status Report 2021 Research-status Report

URL: 

Published: 2021-04-28   Modified: 2025-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi