Updated on 2025/09/30

写真a

 
WASHIO,Takashi
 
Organization
Faculty of Business Data Science Professor
Title
Professor
External link

Research Areas

  • Informatics / Intelligent informatics

Research History

  • Kansai University   Faculty of Business and Commerce   Professor

    2024.10

      More details

Papers

▼display all

MISC

  • AIと固体ナノポアセンサによるウイルス検査

    有馬彰秀, 筒井真楠, 鷲尾隆, 馬場嘉信, 馬場嘉信, 馬場嘉信, 川合知二

    生物工学会誌   101 ( 8 )   2023

  • Measurement Informatics and Its Application in Science Invited

    Takashi Washio

    Proceedings of SciX2022: SciX (The Great SCIentific eXchange) Conference 2022   ( 342 )   2022.10

     More details

    Authorship:Lead author   Language:English   Publishing type:Meeting report  

    researchmap

  • Developments of control system for ion source using machine learning

    Y Morita, M Fukuda, T Yorita, H Kanda, K Hatanaka, T Saitou, H Tamura, Y Yasuda, T Washio, Y Nakashima, M Iwasaki, H W Koay, K Takeda, T Hara, T H Chong, H Zhao

    Journal of Physics: Conference Series   2244 ( 1 )   012105 - 012105   2022.4

     More details

    Publisher:IOP Publishing  

    Abstract

    Various factors influence each other in an ion source. Therefore, when operating an ion source, it is necessary to optimize and adjust various parameters. This time, we performed an experiment to automize adjustment that maximizes the brightness of the beam using machine learning. By automatically adjusting 4 parameters, we succeeded in finding a point with a beam brightness of 4.32 × 10<sup>-6</sup> mA/(imm mrad)<sup>2</sup> in 25 steps. This shows that automatic adjustment using Bayesian optimization is feasible.

    DOI: 10.1088/1742-6596/2244/1/012105

    researchmap

    Other Link: https://iopscience.iop.org/article/10.1088/1742-6596/2244/1/012105/pdf

  • 3D integrated nanopore for single cell lysis to single-molecule DNA detections

    筒井真楠, 横田一道, 有馬彰秀, 鷲尾隆, 馬場嘉信, 川合知二

    応用物理学会春季学術講演会講演予稿集(CD-ROM)   69th   2022

  • Early detection of brain tumors by using next-gen machinery deep learning algorithm-based urinary liquid biopsy

    夏目敦至, 夏目敦至, 安井隆雄, 安井隆雄, 鷲尾隆, 北野詳太郎, 青木恒介, 市川裕樹, 水沼未雅, 高山和也, 高山和也, 齋藤竜太, 若林俊彦, 馬場嘉信, 馬場嘉信

    日本脳腫瘍学会学術集会プログラム・抄録集   39th   2021

  • ナノポアデバイスを用いた単一生体粒子分析

    有馬彰秀, 筒井真楠, 吉田剛, 横田一道, 立松健司, 山崎智子, 黒田俊一, 谷口正輝, 鷲尾隆, 川合知二, 馬場嘉信, 馬場嘉信, 馬場嘉信

    Molecular Electronics and Bioelectronics   31 ( 2 )   2020

  • ナノポアデバイスを用いた単一生体粒子分析—応用物理学会 有機分子・バイオエレクトロニクス分科会研究会 ここまで進んだ有機分子・バイオエレクトロニクス研究

    有馬 彰秀, 筒井 真楠, 吉田 剛, 横田 一道, 立松 健司, 山﨑 智子, 黒田 俊一, 谷口 正輝, 鷲尾 隆, 川合 知二, 馬場 嘉信

    Molecular electronics and bioelectronics = 応用物理学会,有機分子・バイオエレクトロニクス分科会会誌 / 応用物理学会有機分子・バイオエレクトロニクス分科会 編   31 ( 2 )   93 - 96   2020

     More details

    Language:Japanese   Publisher:応用物理学会有機分子・バイオエレクトロニクス分科会  

    CiNii Books

    researchmap

  • 機械学習と分子認識ナノポアを用いた1ウイルス識別

    筒井真楠, 有馬彰秀, 吉田剛, 横田一道, 殿村渉, 谷口正輝, 鷲尾隆, 川合知二, HARLISA Ilva Hanun, 田中祐圭, 大河内美奈

    応用物理学会春季学術講演会講演予稿集(CD-ROM)   66th   2019

  • 外部摂動イオン電流による薬剤耐性大腸菌の識別

    吉川碧海, 安井隆雄, 安井隆雄, 安井隆雄, 嶋田泰介, 嶋田泰介, 山崎聖司, 西野邦彦, 柳田剛, 長嶋一樹, 鷲尾隆, 川合知二, 馬場嘉信, 馬場嘉信, 馬場嘉信

    日本化学会春季年会講演予稿集(CD-ROM)   99th   2019

  • ナノバイオデバイスと機械学習の融合による多項目ウイルス識別

    有馬彰秀, 有馬彰秀, 筒井真楠, 殿村渉, 横田一道, 安井隆雄, 安井隆雄, 嶋田泰佑, 嶋田泰佑, 山崎智子, 立松健司, 黒田俊一, 谷口正輝, 鷲尾隆, 川合知二, 馬場嘉信, 馬場嘉信, 馬場嘉信

    化学とマイクロ・ナノシステム学会研究会講演要旨集   39th   2019

  • フラジェリン認識ペプチド修飾ポアセンサによる大腸菌の個別計測

    大河内美奈, 大河内美奈, 田中祐圭, 田中祐圭, 丸井貴皓, 丸井貴皓, 筒井真楠, 筒井真楠, 横田一道, 横田一道, 鷲尾隆, 鷲尾隆, 谷口正輝, 谷口正輝

    電気化学秋季大会講演要旨集(CD-ROM)   2018   2018

  • フラジェリン認識ペプチドを修飾したポアセンサによる微生物の検出

    大河内美奈, 大河内美奈, 田中祐圭, 田中祐圭, 丸井貴皓, 丸井貴皓, 筒井真楠, 筒井真楠, 横田一道, 横田一道, 鷲尾隆, 鷲尾隆, 谷口正輝, 谷口正輝, 河合知二, 河合知二

    化学とマイクロ・ナノシステム学会研究会講演要旨集   37th   2018

  • Image Reconstruction for Super Resolution Microscope Using Recursive Bayesian Computation

    KIDO Shunsuke, WASHIO Takashi, WAZAWA Tetsuichi, NAGAI Takeharu

    Proceedings of the Annual Conference of JSAI   2018 ( 0 )   3L203 - 3L203   2018

     More details

    Language:Japanese   Publisher:The Japanese Society for Artificial Intelligence  

    &lt;p&gt;We propose to apply Recursive Bayesian Computation to image estimation of SPoD-ExPAN microscopy.The method does not need derivatives of the optimality measure and is supposed to derive images globally better than those of gradient dissent based approaches.In this paper, we present an implementation of the Recursive Bayesian Computation by using Kernel density estimation.Moreover, we introduce regularization to the estimation, and experimentally compare its performance with the case without the regularization.&lt;/p&gt;

    researchmap

  • ナノポア計測と機械学習によるインフルエンザウイルス識別

    筒井真楠, 有馬彰秀, 吉田剛, 横田一道, 殿村渉, 谷口正輝, 鷲尾隆, 川合知二, ILVA Harlisa, 田中祐圭, 大河内美奈

    日本化学会春季年会講演予稿集(CD-ROM)   98th   2018

  • 低アスペクト比ポアセンサと機械学習法による1粒子形状識別 (M&BE研究会 有機分子・バイオエレクトロニクスの最新動向と応用展開)

    筒井 真楠, 谷口 正輝, 鷲尾 隆, 川合 知二

    Molecular electronics and bioelectronics = 応用物理学会,有機分子・バイオエレクトロニクス分科会会誌   28 ( 2 )   65 - 68   2017.5

     More details

    Language:Japanese   Publisher:応用物理学会有機分子・バイオエレクトロニクス分科会  

    CiNii Books

    researchmap

  • ナノポアと機械学習による1細菌の識別 : 物理計測と機械学習で分子認識能を創る

    谷口 正輝, 鷲尾 隆, 川合 知二

    化学 = Chemistry   72 ( 2 )   33 - 38   2017.2

     More details

    Language:Japanese   Publisher:化学同人  

    CiNii Books

    researchmap

  • 微生物結合ペプチドを修飾したポアセンサにおけるイオン電流の応答解析

    服部翔太, イルファ ハヌンハルリサ, 丸井貴皓, 田中祐圭, 大河内美奈, 有馬彰秀, 筒井真楠, 谷口正輝, 鷲尾隆, 川合知二

    化学工学会年会研究発表講演要旨集(CD-ROM)   82nd   2017

  • 微生物判別デバイスの開発に向けたペプチド探索

    服部翔太, 服部翔太, 田中祐圭, 田中祐圭, 有馬彰秀, 有馬彰秀, 筒井真楠, 筒井真楠, 谷口正輝, 谷口正輝, 鷲尾隆, 鷲尾隆, 川合知二, 川合知二, 大河内美奈, 大河内美奈

    化学とマイクロ・ナノシステム学会研究会講演要旨集   35th   2017

  • 遺伝子工学的に開発した蛍光プローブによる細胞生理機能超解像イメージング

    和沢 鉄一, 新井 由之, 河原 吉伸, 中野 雅裕, 松田 知己, 鷲尾 隆, 永井 健治

    人工知能学会全国大会論文集   2017 ( 0 )   2I4OS10b4 - 2I4OS10b4   2017

     More details

    Language:Japanese   Publisher:一般社団法人 人工知能学会  

    &lt;p&gt;近年の超解像蛍光顕微鏡技術開発により、電子顕微鏡に頼らずに細胞内の微細構造の可視化が可能になった。しかし、細胞挙動の理解を今後進めるには、細胞内の構造のみならず、情報伝達物質の動き、温度、酵素反応等の生理パラメータを超解像イメージングで計測し、そこから潜在的要因の検出や状態の推定をすることが必要になってくる。本論文では、蛍光プローブ開発と生理機能超解像イメージングへ向けた取り組みについて発表する。&lt;/p&gt;

    researchmap

  • PC chairs’ preface

    James Bailey, Latifur Khan, Takashi Washio

    Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)   9652   v   2016

     More details

    Language:English   Publisher:Springer Verlag  

    Scopus

    researchmap

  • Anomaly Detection of Informatical Quantum States by Using Machine Learning

    Takashi Washio, The Institute of Scientific and Industrial Research Osaka University

    30 ( 2 )   217 - 223   2015.3

     More details

  • Markowitz Portfolio Selection with Change-point Detection and Sparse Estimation

    29   1 - 4   2015

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Modeling from Big Data

    WASHIO Takashi

    Systems, control and information   58 ( 1 )   3 - 8   2014

     More details

    Language:Japanese   Publisher:Institute of Systems, Control and Information Engineers  

    CiNii Books

    researchmap

  • Abnormal Event Detection in Crowded Scenes using Structured Sparse Learning

    28   1 - 4   2014

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Quantum state data mining

    Ono Takafumi, Okamoto Ryo, Takeuchi Shigeki, Hara Satoshi, Washio Takashi

    Meeting abstracts of the Physical Society of Japan   68 ( 2 )   147 - 147   2013.8

     More details

    Language:Japanese   Publisher:The Physical Society of Japan (JPS)  

    CiNii Books

    researchmap

  • Parametric Min-Cuts for Structure Sparse PCA

    88   109 - 112   2013.1

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Principal Component Analysis using Structured Sparsity via Graph Cuts

    27   1 - 4   2013

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Weighted Likelihood Policy Search

    UENO Tsuyoshi, HAYASHI Kohei, WASHIO Takashi, KAWAHARA Yoshinobu

    112 ( 279 )   165 - 170   2012.10

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    Reinforcement learning (RL) methods based on direct policy search (DPS) have been actively discussed to achieve an efficient approach to complicated Markov decision processes (MDPs). Although they have brought much progress in practical applications of RL, there still remains an open problem in DPS related to model selection for the policy. In this paper, we propose a new DPS method, weighted likelihood policy search (WLPS), where a policy is efficiently learned through the weighted likelihood estimation. WLPS naturally connects DPS to the statistical inference problem and thus various sophisticated techniques in statistics can be applied to DPS problems directly. Hence, by following the idea of the information criterion, we develop a new measurement for model comparison in DPS based on the weighted log-likelihood.

    CiNii Books

    researchmap

  • Structure Learning for Anomaly Localization

    HARA Satoshi, WASHIO Takashi

    112 ( 279 )   17 - 22   2012.10

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    In this paper, we propose a graphical model learning algorithm for an anomaly localization. We introduce a new regularization term that penalizes the row/column-wise difference between two precision matrices and formulate the task as a convex optimization problem. We further provide an optimization algorithm based on an alternating direction method. The validity of the proposed method is presented through a simulation using a real world data.

    CiNii Books

    researchmap

  • Editor's Introduction to "Discrete Structure Manipulation Systems-The Art of Algorithms for Intelligent Information Processing"

    WASHIO Takashi, Takashi Washio

    27 ( 3 )   231 - 231   2012.5

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Recent Development of Intelligent Information Processing with Submodularity(<Special Issue>Discrete Structure Manipulation Systems-The Art of Algorithms for Intelligent Information Processing)

    KAWAHARA Yoshinobu, NAGANO Kiyohito, WASHIO Takashi, Yoshinobu Kawahara, Kiyohito Nagano, Takashi Washio, The Institute of Scientific and Industrial Research (ISIR) Osaka University:Japan Science and Technology Agency (JST), Institute of Industrial Science The University of Tokyo, The Institute of Scientific and Industrial Research (ISIR) Osaka University:Japan Science and Technology Agency (JST)

    27 ( 3 )   252 - 260   2012.5

     More details

  • Path Integral Control on Manifold

    26   1 - 4   2012

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Sparse Inverse Covariance Selection via DAL-ADMM

    26   1 - 4   2012

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Bootstrapping confidence intervals in linear non-Gaussian causal model (人工知能学会全国大会(第26回)文化,科学技術と未来) -- (機械学習)

    Thamvitayakul Kittitat, 清水 昌平, 鷲尾 隆

    人工知能学会全国大会論文集   26   1 - 3   2012

     More details

    Language:English   Publisher:人工知能学会  

    We consider the problem of finding significant connection strengths of variables in a linear non-Gaussian causal model called LiNGAM. In our previous work, bootstrapping confidence intervals of connection strengths were simultaneously computed in order to test their statistical significance. However, such a naive approach raises the multiple comparison problem which many directed edges are likely to be falsely found significant. Therefore, in this study, we tested two multiple testing correction approaches, Bonferroni correction and Mandel's approach, then evaluated their performance. We found that both Bonferroni correction and Mandel's approach are able to eliminate some of falsely found directed edges.

    DOI: 10.11517/pjsai.jsai2012.0_4b1r24

    CiNii Books

    researchmap

  • A feature selection method based on randomized algorithm

    26   1 - 4   2012

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • トクシュウ 「 ベイジアンネットワーク ト ソノ オウヨウ 」 オヨビ イッパン

    83   63 - 70   2011.11

  • Prismatic Algorithm for Discrete D.C. Programming Problem

    KAWAHARA Yoshinobu, WASHIO Takashi

    Advances in Neural Information Processing Systems 24: 25th Annual Conference on Neural Information Processing Systems 2011. Proceedings of a meeting held 12-14 December 2011(NIPS)   111 ( 275 )   93 - 98   2011.11

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    In this paper, we propose the first exact algorithm for minimizing the difference of two submodular functions (D.S.), i.e., the discrete version of the D.C. programming problem. The developed algorithm is a branch-and-bound-based algorithm which responds to the structure of this problem through the relationship between submodularity and convexity. The D.S. programming problem covers a broad range of applications in machine learning because this generalizes any set-function optimization. We empirically investigate the performance of our algorithm, and illustrate the difference between exact and approximate solutions respectively obtained by the proposed and existing algorithms in feature selection.

    CiNii Books

    researchmap

  • A Method for Estimating Binary Data Generating Process

    INAZUMI Takanori, WASHIO Takashi, SHIMIZU Shohei, SUZUKI Joe, YAMAMOTO Akihiro, KAWAHARA Yoshinobu

    111 ( 275 )   155 - 162   2011.11

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    In our previous study, we proposed a method to identify a data generation process governing its given binary data set. However, statistics used in the method were not optimal. In this paper, we report a preliminiary result bu using more proper statistics. The experimental evaluation shows promising performance.

    CiNii Books

    researchmap

  • Learning a Graphical Structure with Clusters

    HARA Satoshi, WASHIO Takashi

    111 ( 275 )   19 - 24   2011.11

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    In this paper, we propose an estimation technique of a graphical model with some unknown clusters. We introduce a new regularization term based on a graph Laplacian matrix which captures the cluster structure. We confirm the validity of the proposed method through numerical simulation.

    CiNii Books

    researchmap

  • 離散データの因果の同定 : 2値から、多値への一般化について—情報論的学習理論と機械学習

    鈴木 譲, 清水 昌平, 鷲尾 隆

    電子情報通信学会技術研究報告 = IEICE technical report : 信学技報   111 ( 275 )   207 - 212   2011.11

     More details

    Language:Japanese   Publisher:東京 : 電子情報通信学会  

    CiNii Books

    researchmap

    Other Link: https://ndlsearch.ndl.go.jp/books/R000000004-I023346021

  • Mining High Dimensional Data in the Info-plosion Era

    WASHIO Takashi

    The Journal of the Institute of Electronics, Information and Communication Engineers   94 ( 8 )   679 - 683   2011.8

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    CiNii Books

    researchmap

  • Learning an Invariant Substructure of Multiple Graphical Gaussian Models

    HARA Satoshi, WASHIO Takashi

    IEICE technical report   110 ( 476 )   177 - 181   2011.3

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    Dependency structure among variables is closely tied to an underlying data generating process. Therefore, learning a dependency structure from observations is an important task in data mining. Especially when multiple data sources involve common substructure among their dependency structures, it implies an existence of an underlying fundamental mechanism. In this paper, we propose a learning algorithm for finding such a common dependency structure from multiple datasets in the case of Gaussian Graphical Model (GGM). Our proposed algorithm is based on a block coordinate descent method and is a natural extension of an existing learning algorithm for GGM. We show the validity of our approach in a numerical simulation.

    CiNii Books

    researchmap

  • ANALYZING RELATIONSHIPS BETWEEN CTARMA AND ARMA MODELS

    25   1 - 4   2011

     More details

    Language:English  

    CiNii Books

    researchmap

  • Simultaneous Learning of Graphical Structures

    25   1 - 4   2011

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Analyzing Optimal Marketing Strategies Over Customers' Networks

    25   1 - 4   2011

     More details

    Language:English  

    CiNii Books

    researchmap

  • Experimental evaluation of a method to estimate the data generating process of a binary variable causal model

    稲積 孝紀, 鷲尾 隆, 清水 昌平

    人工知能学会全国大会論文集   25   1 - 4   2011

     More details

    Language:Japanese   Publisher:人工知能学会  

    DOI: 10.11517/pjsai.jsai2011.0_2e36

    CiNii Books

    J-GLOBAL

    researchmap

  • DNAシーケンスアラインメント手法を応用したスーパーマーケットにおける顧客動線分類に関する研究

    市川昊平, IP Edward Hak-Sing, 矢田勝俊, 鷲尾隆

    人工知能学会知識ベースシステム研究会資料   91st   2011

  • Relational Data Mining on Causal Relations Between Variables

    WASHIO Takashi

    110 ( 76 )   5 - 5   2010.6

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    Under development of sensing techniques to ease simultaneous measurements of many variables and features on objects, the need increases to systematically understand an objective system underlying data generation process where the data is generated under influences from some variables/relations to the other variables/relations. While graphical modeling techniques and time series analyses have been used for this need in the long past period, novel techniques are emerging in recent years. Members in our laboratory currently work on development of novel relational data mining methods focusing on data generation process based on the causality between many variables/elements by using graph mining, statistical causal inference and optimization algorithm. In this talk, we introduce these studies and their application examples and further prospect the future possibility of relational data mining for understanding the data generation process.

    CiNii Books

    researchmap

  • 高次元確率空間における高精度期待値ベイズ推定の検討

    MATSUDA SHUJI, HON NGUYEN HA, WASHIO TAKASHI, KAWAHARA YOSHINOBU, SHIMIZU SHOHEI, INOKUCHI AKIHIRO

    人工知能学会全国大会論文集(CD-ROM)   24th   ROMBUNNO.1A1-4 - 4   2010

  • Issues of statistical large scale causal inference and its challenge based on non-Gaussianity

    75   33 - 36   2009.11

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Identification of exogenously expressed genes by applying independent component analysis

    十河 泰弘, 清水 昌平, 鷲尾 隆

    人工知能学会全国大会論文集   23   1 - 4   2009

     More details

    Language:Japanese   Publisher:人工知能学会  

    DOI: 10.11517/pjsai.jsai2009.0_2c13

    CiNii Books

    J-GLOBAL

    researchmap

  • Preface: Featured section on data-mining and statistical science

    Tomoyuki Higuchi, Takashi Washio

    ANNALS OF THE INSTITUTE OF STATISTICAL MATHEMATICS   60 ( 4 )   697 - 698   2008.12

     More details

  • Research Strategy to Present Your Papers in Prestigious International Conferences(<Special Issue>Writing Good Research Papers for International Conferences)

    WASHIO Takashi, Takashi Washio, The Institute for Scientific and Industrial Research Osaka University

    Journal of Japanese Society for Artificial Intelligence   23 ( 3 )   362 - 366   2008.5

     More details

  • Modeling Dynamics of Massive Dimensional Systems

    NGUYEN Viet Phuong, WASHIO Takashi

    70   239 - 240   2008.3

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Data Intensive Computing : No.1 Discrete Structure Mining(<Lecture Series>Intelligent Computing and Related Issues (1))

    WASHIO Takashi, Takashi Washio, The Institute of Scientific and Industrial Research (ISIR) Osaka University.

    Journal of Japanese Society for Artificial Intelligence   22 ( 2 )   263 - 271   2007.3

     More details

  • Editor's Introduction to "Data Mining and Statistical Science"(<Special Issue>Data Mining and Statistical Science)

    WASHIO Takashi, Takashi Washio, The Institute of Scientific and Industrial Research (ISIR) Osaka University.

    22 ( 2 )   272 - 272   2007.3

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • 知識発見から知識体系発見へ(<特集>編集委員2007年の抱負)

    鷲尾 隆, Takashi Washio

    人工知能学会誌 = Journal of Japanese Society for Artificial Intelligence   22 ( 1 )   22 - 22   2007.1

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • DryadeによるGene Network DAGデータからの飽和頻出木マイニング

    ターミエ アレックサンドル, 鷲尾 隆, 樋口 知之, 玉田 嘉紀, 井元 清哉, 大原 剛三, 元田 浩

    人工知能学会全国大会論文集   6 ( 0 )   7 - 7   2006

     More details

    Language:Japanese   Publisher:一般社団法人 人工知能学会  

    本研究ではバイオインフォマティクスのデータから飽和頻出木をマイニングすることを試みる。対象データの構造はDAGなので、我々のツリーマイニングアルゴリズムDryadeをDAGに適用可能なように改良した。実験でこの効果を確かめる。

    DOI: 10.11517/pjsai.JSAI06.0.7.0

    researchmap

  • Basics and Present of Graph-based Data Mining

    WASHIO Takashi

    IPSJ Magazine   46 ( 1 )   20 - 26   2005.1

     More details

    Language:Japanese   Publisher:Information Processing Society of Japan (IPSJ)  

    CiNii Books

    researchmap

  • (1)Deta Mining Applications : Overview and Prospect(Commentary Series : The Voice of Practitioners in Data Mining)

    WASHIO Takashi, Takashi Washio, The Institute of Scientific and Industrial Research Osaka University

    Journal of Japanese Society for Artificial Intelligence   19 ( 3 )   373 - 375   2004.5

     More details

  • Trend of Data Mining Research and Issues in Application to Pattern Recognition : Let's Work Hard Together

    WASHIO Takashi

    Technical report of IEICE. PRMU   103 ( 295 )   115 - 120   2003.9

     More details

    Language:Japanese   Publisher:The Institute of Electronics, Information and Communication Engineers  

    Data mining is a, technology attracting many researchers and practitioners in recent years. It. is a synthetic technology of various elemetary theories and techniques for data analysis. This fact causes some difficulties to understand data mining, and prevents its collaborative synthesis with the other thechnologies such as pattern recognition. In this note, the state of the art, the recent trend, the issues of data mining are briefly reviewed, and some problems caused by the differences of objectives, scopes and terminologies between data mininig and pattern recognition are discussed.

    CiNii Books

    researchmap

  • A Proposal on Modelling and Its Application of Complex and Social Systems by Scale Constraints

    NIWA Yuji, WASHIO Takashi, MOTODA Hiroshi

    Correspondences on Human Interface   vol.4,No.2,pp.1-8 ( 2 )   1 - 8   2002

     More details

    Language:Japanese   Publisher:Human Interface Society  

    Current difficulty to formulate complex systems may be resolved by introducing scale constraints thanks to the development of advanced information technology. This paper concerns the identification of the first principle equation that governs complex system beviour by applying machine learning and trial study of the relationship between the public affinity and earthquake risk. The basic research of the discovery of the first principle equation has been made in the area of artificial intelligence, a kind of engineering. This outcome was extended to socio-psychology. Thus the framework is considered to be a symbiosis of natural and cultural sciences.

    DOI: 10.11184/hisrm.4.2_1

    CiNii Books

    researchmap

    Other Link: https://jlc.jst.go.jp/DN/JALC/10015804463?from=CiNii

  • Data Mining Contests:Present and Future of Data Mining in Businesss

    WASHIO Takashi

    IPSJ Magazine   42 ( 5 )   467 - 471   2001.5

     More details

    Language:Japanese   Publisher:Information Processing Society of Japan (IPSJ)  

    CiNii Books

    researchmap

    Other Link: http://id.nii.ac.jp/1001/00063806/

  • Mathematical Models in Law Equation Discovery(Special Issue : "Mathematical Models in Artificial Intelligence toward 21st Century")

    WASHIO Takashi, Takashi Washio, Institute of Scientific and Industrial Research Osaka University

    Journal of Japanese Society for Artificial Intelligence   16 ( 2 )   245 - 248   2001.3

     More details

  • A proposal of design reasoning mode that takes notice of design knowledge (the Forth report) : Implementation of a reasoning model in design and its verification

    Nomaguchi Yutaka, Tsumaya Akira, Yoshioka Masaharu, Washio Takashi, Takeda Hideaki, Murakami Tamotsu, Tomiyama Tetsuo

    The Proceedings of Design & Systems Conference   2001 ( 0 )   285 - 288   2001

     More details

    Language:Japanese   Publisher:The Japan Society of Mechanical Engineers  

    Since design knowledge plays a crucial role in design process, we are conducting research on design reasoning model that takes notice of design knowledge. To make a design reasoning model, we have already proposed a reasoning framework of design and analyzed basic operations for the framework, and then propose a synthesis language for describing design operational knowledge. In this paper, we propose our implementation of this model, and verify it to compare design processes replayed in the system with ones of actual design sessions.

    DOI: 10.1299/jsmedsd.2001.10.285

    CiNii Books

    researchmap

  • A proposal of design reasoning model that takes notice of design knowledge (the Third Report) : A proposal of a synthesis language for describing design operational knowledge

    YOSHIOKA Masaharu, Takeda Hideaki, WASHIO Takashi, TOMIYAMA Tetsuo

    The Proceedings of Design & Systems Conference   2001 ( 0 )   281 - 284   2001

     More details

    Language:Japanese   Publisher:The Japan Society of Mechanical Engineers  

    Since design knowledge plays a crucial role in design process, we are conducting research on design reasoning model that takes notice of design knowledge. To make a design reasoning model, we have already proposed a reasoning framework of design and analyzed basic operations for the framework. However, since these operations are just fragments of whole design process, it is necessary to describe design operational knowledge that can control the entire reasoning operations by constructing the sequence of these fragments. So, in this paper, we propose a synthesis language for describing design operational knowledge.

    DOI: 10.1299/jsmedsd.2001.10.281

    CiNii Books

    researchmap

  • History and Perspective of Mining Techniques for Structured Data

    WASHIO Takashi

    14   93 - 96   2000.7

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Mathematical Models in Law Equation Discovery

    WASHIO Takashi

    14   32 - 33   2000.7

     More details

    Language:Japanese  

    CiNii Books

    researchmap

  • Derivation of Exogenously-Driven Causality Based on Physical Laws

    Takashi Washio, Nuclear Reactor Laboratory Massachusetts Institute of Technology

    5 ( 4 )   482 - 491   1990.7

     More details

    Language:Japanese  

    CiNii Books

    researchmap

▼display all

Research Projects

  • Development of Deep Machine Learning Method for Generalized State Space Models Using Prior Knowledge Constraints and Weak Learning

    Grant number:23H00471  2023.4 - 2026.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (A)

      More details

    Grant amount:\46670000 ( Direct Cost: \35900000 、 Indirect Cost:\10770000 )

    researchmap

  • Study on fast and accurate classifier learning method from unlabeled big data

    Grant number:20K21815  2020.7 - 2023.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Challenging Research (Exploratory)

      More details

    Grant amount:\6370000 ( Direct Cost: \4900000 、 Indirect Cost:\1470000 )

    researchmap

  • Research on Principle and Methods of Large-scale Causal Infrerence Based on Nonlinearity

    Grant number:17K00305  2017.4 - 2020.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (C)

    Washio Takashi

      More details

    Grant amount:\4680000 ( Direct Cost: \3600000 、 Indirect Cost:\1080000 )

    There is an increasing need to understand the mechanism of large-scale systems by analyzing big data by statistical causal inference. However, its practical principles and methods have been established only for large-scale systems that are linear and have non-Gaussian noise. This research achieved (1) establishment of a new principle for estimating the causal relationship between many observation variables in a non-linear system with high accuracy, (2) development of statistical causal inference methods for large-scale systems by further extending the new principle, (3) basic performance verification using large-scale artificial data, and (4) practical performance verification using real-world data. We developed practical principles and methods for a wide range of large-scale nonlinear systems though these studies. Furthermore, we presented these results in major international conferences and international journals, and spread the breakthrough method of statistical causal reasoning.

    researchmap

  • 機械学習と最先端計測技術の融合深化による新たな計測・解析手法の展開

    2016 - 2021

    科学技術振興機構  戦略的な研究開発の推進 戦略的創造研究推進事業 CREST 

    鷲尾 隆

      More details

    最先端の計測・デバイス技術と融合した新たな機械学習技術を確立・深化し、従来限界を超える現象・精度の計測実現を目指します。特に計測を念頭とし、データ特徴量抽出手法、事前知識を活かす少数データ推定手法、複数情報源統合推定手法、計測過程を反映した機械学習手法などを開発します。具体的テストベッドとして、先端的ナノギャップナノポアによる高効率、低コストな第4世代DNAシーケンシング技術の確立を取り上げます。

    researchmap

  • Research on Fundamental Algorithms of Discrete Structure Manipulation Systems

    Grant number:15H05711  2015.5 - 2020.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (S)

    MINATO Shin-ichi

      More details

    Grant amount:\134420000 ( Direct Cost: \103400000 、 Indirect Cost:\31020000 )

    In this project, we aimed to construct the core algorithms for discrete structure manipulation systems, and to provide efficient software tools for many researchers in various application areas. Our achievements include that (1) we first succeeded in enumerating all connected sub-block patterns (in total 109.8 billion patterns) of 47 prefectures in Japan, the data is open for all Japanese citizens from the governmental statistics center, and (2) we produced many academic papers, accepted at the top-conferences such as AAAI, WWW, KDD, INFOCOM, AISTATS, SDM, etc.

    researchmap

  • Developement of sensor hardware/software toward breath diagnostics based on multi-dimensional data analysis algorithm

    Grant number:15H03588  2015.4 - 2018.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    Yoshikawa Genki, Washio Takashi, Shiba Kota

      More details

    Grant amount:\14820000 ( Direct Cost: \11400000 、 Indirect Cost:\3420000 )

    Accurate identification of complex smell/odor (e.g. exhaled breath) composed of diverse molecules requires optimization of both hardware (multiple sensors with diverse chemical selectivity) and software (multidimensional data analysis). In this study, sensor system components including receptor materials have been developed together with the detailed investigation into basic analytical methods of sensing data. Further, significant enhancement of predication accuracy of complex smell/odor has been demonstrated through the optimization of sensors based on machine learning of multidimensional sensing data. These studies provide a guideline for hardware-software mutual optimization of sensors.

    researchmap

  • Model Mining: Exploration of search and enumeration methods of local models from super-high dimensional data

    Grant number:26540116  2014.4 - 2016.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Challenging Exploratory Research

    Washio Takashi, SHIMIZU Shohei, KAWAHARA Yoshinobu

      More details

    Grant amount:\3510000 ( Direct Cost: \2700000 、 Indirect Cost:\810000 )

    This study aimed at the exploration of model mining principles, which enable fast search of candidate models representing sub-processes embedded in super-high dimensional and large scale data, and their implementations into some algorithms for applying to experimental problems including medical fields. We established novel principles of random sub-sampling and ensemble modeling for fast and accurate model mining from the large scale data, and developed the methods of half-space mass and and mass based similarity measures by implementing the principles. Finally, by applying these methods to heart disease data in medicine, we succeeded to mine a model of a occurrence mechanism of the heart disease. These outcomes have been presented in Machine Learning:the world top journal of machine learning, ICDM: the world top international of data mining and a major medical journal.

    researchmap

  • Development and Application of Statistical Estimation and Simulation for Super High Dimensional Data Space

    Grant number:25240036  2013.4 - 2017.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (A)

    Washio Takashi, SHIMIZU Shohei, KAWAHARA Yoshinobu, INOGUCHI Akihiro, Ting Kai Ming

      More details

    Grant amount:\45760000 ( Direct Cost: \35200000 、 Indirect Cost:\10560000 )

    In this study, we aimed to develop (1) generic and robust principles of statistical estimation and scenario generation against super high dimensionality, (2) statistical estimation methods using super high dimensional data, (3) probabilistic scenario generation methods for super high dimensional space, (4) an application of these developed methods and simulation techniques, and (5) an international research community.
    Throughout this project, we developed techniques of similarity measure, density evaluation, robust estimation, scenario search, retrieval and clustering, classification, anomaly detection, rare scenario generation, and frequent pattern derivation. We also organized two international conferences and seven international workshops/seminars.

    researchmap

  • Bayesian network structure learning when discrete and continuous variables are present.

    Grant number:24500172  2012.4 - 2016.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (C)

    Suzuki Joe, Washio Takashi, Kano Yutaka

      More details

    Grant amount:\5070000 ( Direct Cost: \3900000 、 Indirect Cost:\1170000 )

    We consider Bayesian network structure learning when discrete and continuous variables are present. The problem is rather hard and very few results are available. I particular, we had to assume that each continuous variable is Gaussian and no two discrete variable should be between a continuous variable. In this research, we mathematically prove consistency (the correct structure is estimated as the sample size increases). In particular, we proposed applications to independence testing and estimation of mutual information.

    researchmap

  • Chemical structure mining for adverse reactions and early stage signal detection

    Grant number:24240025  2012.4 - 2016.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (A)

    Okada Takashi

      More details

    Grant amount:\37440000 ( Direct Cost: \28800000 、 Indirect Cost:\8640000 )

    Effective components of drugs are accumulated in PharmCompo database. Each entry in the database has ATC codes. A new algorithm has been proposed to detect classification nodes and component nodes characteristically related to an adverse reaction by drugs. We have applied this algorithm to adverse event reports in JADER using anaphylaxis and other 6 reactions. Drug classifications and component drugs were detected causing frequent reactions. Browsing the structures of these drugs led to several substructures, and the succeeding structure refinement process enabled the proposal of structural alerts to these adverse reactions.

    researchmap

  • Co-adaptive BMI by reinforcement learning based on prediction of users' latent mental states

    Grant number:24300093  2012.4 - 2015.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    KAWANABE Motoaki, KANEMURA Atsunori, UENO Tsuyoshi, WASHIO Takashi

      More details

    Grant amount:\17940000 ( Direct Cost: \13800000 、 Indirect Cost:\4140000 )

    Toward a real-time co-adaptive BMI algorithm for providing flexible feedback schemes based on users' latent mental states, we developed reinforcement learning procedures to construct BMI decoders and representations for brain activities to infer the mental states. For the former topic, based on the weighted likelihood, we establish a theoretical framework to determine the optimal state modeling, namely the dimension of the mental states and their transition rule, to design an appropriate policy model, and to execute reinforcement learning simultaneously. For the latter topic, we proposed various generalizations of the standard feature extraction method CSP (common spatial pattern) to construct robust features against subject-to-subject variability and non-stationarity in brain signals. By integrating these element technologies, we implemented a BMI feedback device with portable EEG and a ball lamp, and tested its usefulness with a few subjects in a real-world environment.

    researchmap

  • Learning Probabilistic Simulation Models for Rare Event/Condition Occurrence

    Grant number:24650069  2012.4 - 2014.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Challenging Exploratory Research

    WASHIO Takashi, IBA Yutaka, SHIMIZU Shohei, KAWAHARA Yoshinobu

      More details

    Grant amount:\3770000 ( Direct Cost: \2900000 、 Indirect Cost:\870000 )

    Various approaches for learning probabilistic models from given data and background knowledge have been studied in the past, however, studies on the probabilistic model learning for rare/special conditions have been very limited. In this study, we developed an efficient and accurate approach to learn probabilistic simulation models for the rare/special conditions by using a given data set and its associated background knowledge. Moreover, we demonstrated a novel framework for the probabilistic estimation and prediction of rare/special events and scenarios through its applications to a rare and large scale natural disaster.

    researchmap

  • Statistical prediction, causation, incomplete data analysis and foundation of sciencee

    Grant number:22300096  2010.4 - 2014.3

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    KANO YUTAKA, DEGUCHI Yasuo, WASHIO Takashi, HAMAZAKI Toshimitsu, TAKAGI Yoshiji, SUGIMOTO Tomoyuki, TAKAI Keiji, NAITO Kanta, SHIMIZU Shohei, KATAYAMA Shota, YAMAMOTO Michio, SONG Xinyuan, JAMSHIDIAN Mortaza, HYVARINEN Aapo, YUAN Ke-hai

      More details

    Grant amount:\17810000 ( Direct Cost: \13700000 、 Indirect Cost:\4110000 )

    Analysis of incomplete data has been troublesome both theoretically and practically. In particular nonignorable missingness has been a serious issue in statistics. An alternative perspective of the theory of missing data analysis is to provide an insightful view of statistical causal inference. Some notable research outcomes include development of the analysis of doubly censored data, a new method of exploring causal structure for data with latent confounders via the LiNGAM approach, incomplete data analysis with a shared-parameter model and development of the EM algorithm with constraints.

    researchmap

  • Establishment of Statistical Estimation Principle for Super HighDimensional Data and Its Application to Large Scale Data Mining

    Grant number:22300054  2010 - 2012

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    WASHIO Takashi, HIGUCHI Tomoyuki, INOKUCHI Akihiro, KAWAHARA Yoshinobu, SHIMIZU Shohei, NAKANO Shinya

      More details

    Grant amount:\17290000 ( Direct Cost: \13300000 、 Indirect Cost:\3990000 )

    Upon analysis of dimensionality curse, we characterized “hyper-sphere concentration effect”, “probability concentration effect” and “sparsity effect”of super high dimensional data, and proposed an accurate and robust estimation method against the former two effects.

    researchmap

  • 高次元観測データからの大規模対象状態に関する未来予測と管理戦略策定手法の開発

    Grant number:21013032  2009 - 2010

    日本学術振興会  科学研究費助成事業  特定領域研究

    鷲尾 隆

      More details

    Grant amount:\4800000 ( Direct Cost: \4800000 )

    本最終年度は,集約された巨視ダイナミクスと粒子シミュレーション最適化に基づいて、(A)操作可能なパラメータによる目標状態へ対象系を導く管理戦略を導出する方法の検討・確立と、(B)前年度最後の予備的実適用を通じて明らかになった当初技術(1)各粒子周りの局所ダイナミクスを逐次導出して次元の呪いの問題を回避しつつ効率的かつ適切に修正すべき状態変数組を探索する下法、(2)各粒子の選択状態変数組について局所ダイナミクスから逐次状態を修正予測する方法、(3)多数粒子の予測状態軌跡を巨視的に集約する方法の問題点克服に取り組んだ。まず、(B)については、粒子群から確率密度推定する際の近似を修正することで、計算量を抑えたまま高い推定精度を確保する方法を確立し、(1)(2)(3)何れの問題点をも解決することに成功した。(A)については、本改良・拡張した手法をRFIDタグチップによる大規模スーパーマーケットの商業物流・人間移動ユビキタス追跡システムデータに適川し、大規模変数次元時系列観測データから得られるダイナミクスモデルに関して妥当な未来予測と有効な管理戦略が策定できるかを例題を通じて評価した。数値実験を繰り返し、管理戦略策定方法の構築と改良を進め、実問題に適川可能な方法論を確立し、当該実問題で有効性を実証した。特にこれを通じ、従来手法で直接推定が不可能であった大規模スーパー店舗における顧客の各売場毎の滞在時間と商品購入確率を高精度推定することを可能にする手法を得た。

    researchmap

  • 不完全データからの大規模半正定行列推定手法の探究と量子情報計算実験推定への応用

    Grant number:21650029  2009 - 2010

    日本学術振興会  科学研究費助成事業  挑戦的萌芽研究

    鷲尾 隆

      More details

    Grant amount:\3000000 ( Direct Cost: \3000000 )

    本研究では,大規模半正定行列で表わされる不完全データから数学的に許容される完全な半正定行列を高精度,高効率に推定する手法を探求した.本最終年度は、(1)前年度に開発した手法を観測誤差と欠測を含むデータに適用し性能検証を実施し、(2)その手法を量子情報計算実験結果データに適用し、量子情報計算装置の実験動作と理論予想との合致判定法の提供を試みた。
    (1)開発手法の観測誤差と欠測を含むデータへの適用による性能検証
    容易に得られる大量データの例として、米国のNational Oceanographic Data Centerにおいて公開されている南太平洋領域の巨視的な海洋波動に関する人工衛星リモートセンシング時系列データを取り上げた。人工的に約半分を削除したデータから波高の推定を行い、原波高データと照合して予測精度の検証を行った。その結果、従来の統計的最尤推定で得た結果に比して、約3倍の精度向上を得ることができた。
    2)量子情報計算実験結果への適用による実験動作と理論予想との合致判定法の開発
    まず、量子情報計算シミュレータを構築し、人工的に実験環境の変化、実験パラメータの変化を導入したシミュレーションデータを作成した。このデータに以上により開発と性能確認が終了した推定手法を適用した。その結果、導入した種々の変化を妥当に反映する推定結果を得た。次に、量子情報計算実験の実データへの当該手法の適用を実施し、実験条件の変化を反映した推定が行えることを確認した。

    researchmap

  • Development of Causal Structure Mining Method for Large Scale Dimensional Data and Construction of Gene Function Knowledge Base

    Grant number:19200013  2007 - 2009

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (A)

    WASHIO Takashi, KANO Yutaka, IMOTO Seiya, OHARA Kouzou, TERMIER Alexandlre, INOKUCHI Akihiro, SHIMIZU Shohei, KAWAHARA Yoshinobu

      More details

    Grant amount:\38480000 ( Direct Cost: \29600000 、 Indirect Cost:\8880000 )

    Scientists attempt to figure out function of each gene through the analysis of causal relations between gene expressions by using measurement data of the many gene expression variables (large scale dimensional data). However, the analysis of causal relations between dozens or hundreds of variables is hardly performed manually. In spite of this problem, the number of variables to which the computer based causal analysis is applicable is limited to 20-30 in the state of the art. Accordingly, this work developed a novel principle of the statistical causal analysis, and furthermore constructed a knowledge base of the functional relations among expressed genes for the scientists by using our developed approach.

    researchmap

  • 大規模次元観測時系列からのダイナミクス知識体系化と理解支援手法の開発

    Grant number:19024048  2007 - 2008

    日本学術振興会  科学研究費助成事業  特定領域研究

    鷲尾 隆, 矢田 勝俊, 大原 剛三, 猪口 明博

      More details

    Grant amount:\6100000 ( Direct Cost: \6100000 )

    本研究では, 大規模次元で表わされる状態が時間的に変化する対象系を観測したデータから, 対象の状態遷移規則を表す知識体系とそれを理解支援する技術を確立し, ICチップなどによる商業用物流・人間移動のユビキタス追跡分析・監視システムを実現する基礎原理を得ることを目的とした.
    今年度は, 前年度の手法適用を通じて問題が明らかになった,(1)個々の状態遷移規則同士の因果関係が従うべき数理的, 確率的, 物理的制約を用い, 対象の有意味な状態遷移に関する知識体系を同定する技術の開発, 及び(2)そこから特定部分状態関係を含む状態遷移規則やその規則同士の特徴的関係を把握する技術の問題点を克服する改良, 拡張に取り組んだ. 前者に関しては対象システムが取る可能性のある多くの状態候補を計算し, それら状態を確率的に統合して対象の状態とその状態遷移を推定する原理が, 特に大規模次元状態空間内で高精度, 高効率に動作する技術を開発した. 後者については, 更に特に実状態である可能性の高い状態を導く特徴的な遷移を把握し, 結果の理解容易性と同時で状態推定精度を高める方法を開発した.
    以上のために, 大阪大学の研究代表者(鷲尾)と関西大学の連携研究者(矢田)間の定期的検討会を持って緊密に連携し, 更に改良・拡張した手法を実データに適用して, 大規模変数次元時系列観測データのダイナミクスに関して総合的な知識体系を得, そのユーザー理解支援を十分に実現可能な技術改良, 拡張を行った. また, この研究過程において, 2名の大阪大学産業科学研究所の研究者(大原, 猪口)から, 主にデータ処理や実験検証の面で連携研究者として協力を得た.

    researchmap

  • Knowledge base construction of motif chemical structures causing various bioactivities

    Grant number:18200010  2006 - 2008

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (A)

    OKADA Takashi, TAKAHASHI Yoshimasa, WASHIO Takashi, FUJISHIMA Satoshi

      More details

    Grant amount:\42900000 ( Direct Cost: \33000000 、 Indirect Cost:\9900000 )

    researchmap

  • 大規模次元時系列の知識発掘・モデル化原理確立と商業ユビキタスデータによる検証

    Grant number:18049052  2006

    日本学術振興会  科学研究費助成事業  特定領域研究

    鷲尾 隆, 大原 剛三

      More details

    Grant amount:\2900000 ( Direct Cost: \2900000 )

    近年の情報ネットワークやセンシング技術の発展により,社会的インフラから逐次出力される重要情報が大規模次元時系列となっている.しかし,従来の統計やデータマイニングで対象とし得る時系列変数は数十次元止まりであった.本研究では,時間軸方向を含めた部分共起分析により,一般的計算機を用いて数万〜数百万次元の時系列からの知識発掘やモデル化を行う基本原理の確立を行った.また,ICタグにより得られる代表的大規模次元時系列である商業物流・人間移動ユビキタス追跡データによる実適用性検証を行った.
    具体的には,従来の統計やデータマイニングの時系列データ解析では,複数時刻のベクトルやトランザクションの関係を決定的または確率的関数Fでモデル化したのに対して,本研究では部分ベクトルや部分トランザクション間の関係Rkを用い,それらを多数総合するE(R1, R2,…, RN)により全体関係を表す方法を提案した.また大規模次元データから効率的かつ完全に部分的関係を導くため,部分共起分析を時間方向に拡張適用した.これにより,一般的計算機を用いて数万〜数百万次元時系列の解析が可能となった.
    更に重要社会インフラであるICチップによる商業用物流・人間移動のユビキタス追跡分析・監視システムを取り上げ,出力される膨大な製品や人間に起こる事象や時間,位置などの大規模次元時系列データへ提案手法を適用し,良好なモデリング性能,知識発掘性能を確認した.
    本研究により,雑誌論文を含む16件の発表成果と著書1件,特許出願1件の成果を得た.

    researchmap

  • 数値相関ルール高速完全探索手法の開発と薬品処方規則発見への適用評価

    Grant number:17650042  2005 - 2007

    日本学術振興会  科学研究費助成事業  萌芽研究

    鷲尾 隆, 大原 剛三, 猪口 明博, 元田 浩

      More details

    Grant amount:\3300000 ( Direct Cost: \3300000 )

    これまで得た原理、アルゴリズムの拡張と実データ適用評価を行なうため、次の3課題を実施した。
    1)定量的相関ルール探索原理の拡張
    これまでの評価結果に基づき、更なる性能の向上を目指した探索原理の拡張検討を行った。
    2)定量的相関ルール探索アルゴリズムの拡張・改良
    上記原理の拡張に伴い、探索アルゴリズムの更なる拡張・改良、計算機実装とその性能評価を継続的に行なった。
    3)上記アルゴリズムの医療治療データを用いた適用評価
    以上で実装された探索アルゴリズムを医療分野の治療データに適用し、実解析を行なった。そして解析結果に基づき、当該アルゴリズムと実装プログラムの速度、得られたルールの質の評価を行った。更に、医療に留まらず、社会アンケート調査、マーケティング分野データへの適用も行なった。これら追加評価実験では、特定分野に限定されない開発手法の一般的有効性の検証を行うことができた。更に、専門医師や社会科学、マーケティング分野の専門家からレビューを受け、発掘された数値相関ルールが、十分に各分野の専門知識の増強、新たな知見の発見に資することを確認した。

    researchmap

  • 計算物理学とデータマイニングの融合による結晶学への現実的・効率的アプローチ

    Grant number:17650039  2005 - 2006

    日本学術振興会  科学研究費助成事業  萌芽研究

    TU BAO HO, 鷲尾 隆, 河崎 さおり, 三谷 忠興, HIEU Chi Dam, 尾崎 泰助, DAM Hieu Chi

      More details

    Grant amount:\3500000 ( Direct Cost: \3500000 )

    本プロジェクトは、情報科学的観点及び概念を材料科学研究に導入し、材料の解析・設計から新規材料開発に至る材料科学研究プロセスの新しい枠組みを提案することをスコープとした。今回、新たな材料科学研究手法の具体的なプロトタイプを行い、その拡張性の検討を行うことを通じて、提案した枠組みが研究手法に転換をもたらす革新的アプローチであることを実証した。本研究プロジェクトでは、結晶学研究に対し最大エントロピー法に物理学の予備的知識の導入、粉末回折スペクトルから極めて正確に結晶の電子密度を予測するスキームの確立、フラーレン材料を対象とする分子性結晶構造解析への適用性の実証、について段階的に取り組み、いくつかの材料研究において成功を収めた。金属ドープフラーレン材料に関しては、高温・高圧下のフラーレンベースネットワーク材料の構造決定に成功した。新規材料設計の研究においては、物理の第一原理計算手法を用いて、ナノテクノロジーに有望なカーボンナノチューブを研究の対象として取り上げ、新規であるカーボンナノチューブに吸着した金属クラスターの電子状態を明らかにし、優れたその触媒機能性のメカニズムを解明した。一方、この研究を支援する情報科学、特にデータマイニング分野では、カーネル手法、データ・発見プロセスおよび規則の可視化、ルール帰納法、最適化に関する基礎研究上の諸課題について、材料科学における実証・応用とともに成果をあげた。材料科学研究においては、既存の計算物理学の適用に加え、本研究を拡張し、大量・複雑なデータに潜在する関係性を発見することで知見を獲得し、材料の構造解析など計算力が要求されるプロセスを準自動化する等、新規材料設計・開発への効率的で効果的な支援手法の確立が期待される。

    researchmap

  • Development of 3 dimensional graph mining techniques and systems to identify physiologically active parts in chemical compounds

    Grant number:16300045  2004 - 2006

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    WASHIO Takashi, OHARA Kouzou

      More details

    Grant amount:\14400000 ( Direct Cost: \14400000 )

    We have established one of the fastest approaches to mine two dimensional graph structures in the world. However, any mining approaches to mine three dimensional graph structures including three dimensional topology, quantitative distance and coordinates have not been explored. The primary goal of this project is to develop a novel three dimensional graph mining technique. The secondary goal is to adapt the technique to identify physiologically active parts in chemical compounds, since the discovery of the physiologically important structures in chemical compounds is the key to find candidate medicine. Moreover, the system for the identification of the physiologically active parts in chemical compounds has been developed.
    The system has been developed in the following two stages.
    (i) Development of a prototype system for each analysis function has been respectively developed, and the performance of each system has been respectively evaluated. These functions are :
    * Comprehensive representation of substructures of three dimensional graphs
    * Function to relate to the chemical molecule orbit computation
    * Function to mine upper layer structures of three dimensional graph structures in large scale molecules
    (ii) Development of a demonstrative and synthesized system to identify physiologically active parts in chemical compounds
    Upon the study in this year, we presented 12 papers, 1 book chapter and 1 patent submission.

    researchmap

  • Development of Constructive Induction Method of Useful Attributes from Complex Structured Data

    Grant number:16300046  2004 - 2005

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    MOTODA Hiroshi, WASHIO Takashi, YOSHIDA Tetsuya, OHARA Kouzou

      More details

    Grant amount:\13800000 ( Direct Cost: \13800000 )

    In data mining where a set of useful knowledge is to be mined from a huge amount of data, the standard practice is to use the original attribute which is used in the original data representation. However, it often happens that the original attributes are not expressive enough and constructing new attributes from the original ones is inevitable. This is called feature construction and yet a better method is to be found. In this research a new feature construction method that is interleaved in the construction of a decision tree is developed and its performance is tested using both artificial and real world datasets. Since the forms of the data to handle become diversified and graph is a good way to represent data of general form, a graph mining method based on sequential chunking method is coupled with a decision tree construction method. The subgraph found at each decision node can be considered as a constructed attribute. The biggest problem of being unable to find overlapping patterns by the straightforward chunking can be avoided by devising pseudo-chunking. The resulting CI-GBI (Chunkingless Graph-based Induction) is now able to do complete search by setting the values for the parameters appropriately. Since it does not use the notion of anti-monotonicity of subgraph subsumption, it can find subgraphs which other state-of-the-art approaches cannot find. Further, because it is guaranteed that the frequency counting of the found subgraphs is accurate, various indices that use frequency, e.g. information gain, are also evaluated accurately and CI-GBI becomes better suited as a feature construction component in decision tree construction. Subgraph search is called recursively during the tree construction and the best feature is constructed on the fly at each decision node. Compared with the straightforward chunking approach, the size of the constructed tree becomes much smaller and the predictive accuracy for an unseen instance becomes better. The application to the chronic hepatitis dataset indicated that it is indeed possible to predict the liver cirrhosis by blood test alone.

    researchmap

  • High dimensional data mining of drug therapeutic effects by the analysis of single nucleotide polymorphism and chemical structure of drugs

    Grant number:14208032  2002 - 2005

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (A)

    OKADA Takashi, ICHIISHI Eiichiro, WASHIO Takashi, OYAMA Mayumi

      More details

    Grant amount:\31980000 ( Direct Cost: \24600000 、 Indirect Cost:\7380000 )

    The objectives of this research project are to develop a methodology of data mining for high dimensional data, and to apply it to the analysis of structure-activity relationships of drugs, the discovery of important genes with SNPs for medical care as well as finding essential factors influencing the quality of patients' life. Various experiments are done to accomplish the research, including wet lab works using DNA microarray. Several results achieved to date are shown below.
    1.Attribute generation system from the chemical structure of drugs is developed to give many linear fragments. The attribute selecting scheme is also introduced to facilitate the effective and efficient mining.
    2.The cascade model, a mining method developed by Okada, was extended to organize rules into principal and relative rules. Topographical expression of rules was also introduced to provide the better understanding of datascape.
    3.The application of the above method to Dopamine receptor ligands resulted in the discovery of agonist and antagonist pharmacophores.
    4.Microarray experiments of insulin-resistant diabetic patients led to the identification relevant SNPs. Further, important genes were discovered, which are evoked by locomotor stimulation, Helicobacter pylori and retinoic acid used to prevent cancer.
    5.Software has been developed to evaluate SNPs effects in receptor-ligand interaction and transcription to RNA.
    6.Clinical records in the hospital for the old are collected and their CYP450 SNPs were investigated. The analysis has shown that pharmaceutical interactions are important for the patients' quality of life.
    7.A method was developed to mine numerical association rules and to utilize them for the classification.
    8.Measurement of fingertip pulse waves was established as a simple and convenient method to analyze the human's psychological conditions.

    researchmap

  • 多様な形式データからの特徴抽出に基づく一元的検索手法の開発

    Grant number:14658102  2002 - 2004

    日本学術振興会  科学研究費助成事業  萌芽研究

    鷲尾 隆, 元田 浩, 吉田 哲也

      More details

    Grant amount:\3400000 ( Direct Cost: \3400000 )

    本年度の研究実績は以下の通りである。
    1.データ形式を超えた検索手法の開発
    前年度までに,画像情報など二次元配列データに関して,二次元のビット配列形式を解いて単なるビット配列に変換し,共通のデータフォーマットを有する特徴量に変換する手法を開発したが,最終年度は画像に限らず,テキスト文書を含む一般の非暗号化バイナリーデータに関して,データから形式依存のビット配列情報を捨象し,残された情報を数学的な不変量に縮約して特徴量に変換する手法を確立した。ビット配列情報から規則順序形式を捨象し一般的なビット配列に変換した.更に数学的不変量を抽出し,検索の手がかりとなる特徴ベクトルを構成した.また,最終年度はデータ形式を超えた高速検索を可能にするべく,被検索データのデータ構造と検索アルゴリズムの開発を行った.特徴ベクトルから高速に情報検索することができるように,いずれの特徴ベクトルがいずれのデータから得られたものであるかを紐付けする逆引きファイルを構成した.そして,検索時には実データを見ることなく逆引きファイル情報を参照することで,高速な検索を可能とした。これにより,種々の構造を有するデータ形式に適用可能な高速検索手法を得た.
    2.検索システムのプロトタイプ作成による性能評価と手法修正
    上記で新たに開発した手法やアルゴリズムをデータサーバ計算機にプログラムとして実装した.性能評価として検索精度及び速度を評価した.その結果,前年度には二次元配列データなどの構造データに関しては数分単位の検索時間が必要とされるたが,最終年度は上記の手法開発により大幅な高速化が図られ,数秒で構造データの検索が可能になった.更に二次元配列構造に限らず,テキストや系列構造,木構造,グラフ構造など,多様な構造データに関して検索性能を検証し,いずれに関しても所与の性質,類似性を持った構造データを高速に検索できることを確認した.
    以上により,本研究の当初の目的である既存のデータ形式に留まらず将来新たに生み出されるであろうデータ形式にも対応しうる,データ内容に共通した不変な数学的特徴を抽出する原理,それによって類似性を判定する原理,及びそれらに基づく検索手法が得られた.

    researchmap

  • 「情報洪水時代におけるアクティブマイニングの実現」の推進と評価

    Grant number:13131101  2001 - 2004

    日本学術振興会  科学研究費助成事業  特定領域研究

    元田 浩, 有川 節夫, 沼尾 正行, 山口 高平, 津本 周作, 鷲尾 隆

      More details

    Grant amount:\36600000 ( Direct Cost: \36600000 )

    研究実績は以下のとおり.
    1.総括班会議,計画研究代表者会議を召集し,各計画研究の進捗度を評価し,必要な助言を与え,対策を講じた.総括班会議は2回,計画研究代表者会議は1回開催した.総括班会議では領域の全体計画を,代表者会議では各計画間にまたがる技術的な細部を議論した。
    2.各計画研究の共通データ解析の進捗状況を把握・評価・助言し必要な対策を講じ成果を実証することを目的に,複数グループ間の会議,共同作業を計34回実施した.
    3.情報処理学会,人工知能学会,電子情報通信学会の関連する研究会との合同研究会にてアクティブマイニングの特集を企画(平成16年12月4〜7日)し,本特定領域の全計画研究から成果を発表した.
    4.人工知能学会誌にアクティブマイニング特集を企画し,Vol.20,No.2に掲載した.また,SpringerよりLecture Note on Artificial Intelligence LNAI3430にてActive Miningの編書を出版した.
    5.第3回アクティブマイニングに関する国際ワークショップ(平成16年6月)を開催し,本特定領域研究から多数の成果を発表した.まだ知識獲得に関する国際ワークショップを開催し,そこでもアクティブマイニングの成果を発表した.
    6.今年度の共通データ解析結果を別に報告書にまとめ刊行した(平成16年12月).さらに,4年間の成果をまとめた報告書を最終成果報告書として刊行した(平成17年3月).また,4年間の成果報告会を兼ねた公開シンポジウムを淡路夢舞台国際会議場にて,本特定領域研究の研究者が全員参加の下に開催した(平成17年2月).

    researchmap

  • 構造データからのアクティブマイニング

    Grant number:13131206  2001 - 2004

    日本学術振興会  科学研究費助成事業  特定領域研究

    元田 浩, 鷲尾 隆, 大原 剛三, TUBAO Ho, 矢田 勝俊, 吉田 哲也

      More details

    Grant amount:\64800000 ( Direct Cost: \64800000 )

    研究実績は以下のとおり.
    1.グラフ構造データからの決定木構築プログラムDT-GBIでの探索過程に領域知識の制約(指定したパターンを含む,含まない)を導入した.
    2.ペアを仮想ノードとして扱い,チャンキングをせず探索する新グラフマイニング手法Cl-GBIを開発した.適切なパラメータ設定により完全探索が可能になり,GBIの数え落としの問題点などを解決した.
    3.上記Cl-GBIを組み込んだ決定木構築プログラムDT-ClGBIを開発し,肝炎データセットで性能を評価した.
    4.数値データを伴うデータから,数値を記号離散化することなしに相関の高い数値区間を自動抽出する原理を確立し,それに基づく数値相関規則導出手法を開発した.
    5.ユーザ指向データマイニングシステムD2MSの肝炎患者に関するルールの理解容易性向上を確認し,多数のルールから統計的に有意なものを選定する手法とルール学習において領域知識を表現の制約に加える手法を提案した,
    6.科学データマイニングとしてゲノムおよび結晶データを並行して解析した.前者に関しては,SVMによるタンパク質の2次構造におけるβターンの予測手法を拡張しγターンを予測した.後者に関しては,粉末回折データから結晶構造を同定する手法を遺伝的アルゴリズムに基づき開発した.
    7.意味的まとまりを捉えたパッセージの集合として文書を表し,トレランス・ラフ集合モデルによるソフトマッチを導入し,意味を反映した相関ルールを得る手法を開発した.
    8.グラフマイニング手法AGMを消費者行動データに適用し,アクティブマイニングによる実証実験を行い新しいデータを収集した.アルコール市場分析から得られた知見に基づき,実際の店舗で店頭プロモーションを行った結果,対象商品の売り上げ増加,関連商品の同時購入頻度の増加を検証することができた.

    researchmap

  • Development of Knowledge Acquisition System that can Adapt to Environment Change

    Grant number:13558034  2001 - 2003

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    MOTODA Hiroshi, SATOH Ken, YOSHIDA Tetsuya, WASHIO Takashi, TERABE Masahiro

      More details

    Grant amount:\13600000 ( Direct Cost: \13600000 )

    In this study an attempt is made to integrate a knowledge acquisition technique which is based on the notion of refinement of existing knowledge introducing the finding from cognitive science and an inductive learning technique which has been developed in the field of machine learning to induce a classifier from accumulated data, to propose a new knowledge acquisition technique to fuse these two different knowledge sources into an operational knowledge, and to verify its effectiveness using real world datasets. More concretely, the following study has been conducted: l) to study a method in which there is no need to know how the knowledge has been acquired and stored in the knowledge base and it is assured that the acquisition of new knowledge does not cause the problem of inconsistency with the existing knowledge, 2) to study a method to conduct continuous knowledge acquisition while automatically identifying which pieces of knowledge have become useless and deleting them still maintaining the overall consistency and the understandability of the constructed knowledge base, 3) to study a method to utilize the accumulated data in such a way that switching between two different knowledge sources (i.e. human exert and accumulated data) can be made at any time of knowledge acquisition without rebuilding the knowledge base from scratch and adapt to environment changes. The developed system has been tested against many datasets of different properties and confirmed to exhibit satisfactory performance. It is now possible to start constructing a knowledge base system acquiring initial pieces of knowledge from human expert and then switching to inductive learning later when abundant data have been accumulated. System developer no more need worry about which pieces of knowledge to delete.

    researchmap

  • DEVELOPMENT OF A GRAPH STRUCTURE DATAMINING METHOD AND IDENTIFICATION SYSTEM OF ACTIVE MOLECULE SUBSTRU CTURES

    Grant number:12480088  2000 - 2002

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    WASHIO Takashi, YOSHIDA Tetsuya, OKADA Takashi, MOTODA Hiroshi

      More details

    Grant amount:\15400000 ( Direct Cost: \15400000 )

    In the first fiscal year, the theoretical framework of graph structure data mining was investigated, and a prototype system for active molecule substructure identification was developed. In this work, the representation of graph structure data in computers and search principle of characteristic graph patterns are studied. Them, the survey of techniques in chemistry which can be introduced to our work has been conducted, and these techniques were reflected in the prototype system. Finally, the basic performance of the prototype system has been evaluated through the substructure extraction in carcinogenetic and mutagenetic chemical component data.
    In the nest fiscal year, the framework of the graph structure data mining was extended to be more efficient in terms of computation time and memory consumption, and the real scale system for active molecule substructure identification has been developed. The algorithm for the efficient computation time and memory consumption was developed, and under the comparison with the conventional techniques in chemistry, the function of the real system was designed. Then, the principle and the algorithm of the real system was modified and extended to enable the graph structure data mining on the massive graph structure data.
    In the final fiscal year, further functions desired to be implemented in the view of chemical analysis were investigated based on the real system developed in the former year, and some functions which can be implemented feasibly were added to the real system. Then, from the view points of the chemical engineering and the computational theory, the practicality and the wide applicability of the real system have been evaluated. Through these evaluations, the practical and high performance of the developed real sysem has been confirmed. The effort to develop commercial system under collaboration with industries is currently underway.

    researchmap

  • Integrated Machine Learning Workbench for Data Mining

    Grant number:11694159  1999 - 2001

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    MOTODA Hiroshi, YOSHIDA Tetsuya, WASHIO Takashi

      More details

    Grant amount:\8700000 ( Direct Cost: \8700000 )

    A new generation of computational techniques and tools is required to support the extraction of useful knowledge from the rapidly growing volumes of data. In this research project we aimed to develop effective methods for feature selection, instance selection and feature construction and integrate them to form a basis of workbench for machine learning and data mining. For feature selection, various performance measures such as distance measure, uncertainty measure, dependency measure, consistency measure and error rate, and various search methods such as heuristic search, complete search and random search were investigated and a design strategy was proposed as to which method to use for which kind of dataset. Further, a new method ABB was proposed that uses consistency measure and performs a very efficient complete search. For instance selection, a new method S^3 Bagging which combines random subsampling and committee learning method was proposed and it was expected that this reduces the amount of data by 90%. For feature construction, two new methods were proposed. One is multi-strategy learning in which graph-base induction GBI that is based on repeated chunking of paired nodes was used as a feature constructor for use in decision tree classifier. Another is to construct new features from association rules. Both were tested against various datasets and conformed effective. All of these are components of the workbench, and we expect that this contributes to mining better knowledge more efficiently.

    researchmap

  • Fundamental theories of ontology and development of an environment for ontology construction

    Grant number:11480076  1999 - 2001

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    MIZOGUCHI Riichiro, SETA Kazuhisa, KITAMURA Yoshinobu, IKEDA Mitsuru, WASHIO Takashi, MOTODA Hiroshi

      More details

    Grant amount:\13200000 ( Direct Cost: \13200000 )

    Although a few convincing ontological theories exist, there is few environment for supporting ontology development based on such a theory. This research has been done aiming at development of fundamental theories for ontology design & construction and that of a support system of an ontology based on that theory. The research is firmly based on the principal investigator's research policy, that is, "content-oriented AI" which is an enterprise of bridging the gap between a theory and practice. The major achievements include:
    (1) theories for is-a and part-of links, a relation, and a role concept. The is-a theory defines a class and an instance by introducing "the intrinsic property" of a thing which helps distinguish between is-a and part-of relations. The theory of a relation introduces "whole concept" and "relational concept" which are two sides of a concrete concept composed of more than one component. The theory of roles is very useful to conceptualize domain concepts which are full of roles.
    (2) Development of guidelines for identifying role concepts in the world of interest. The guideline is based on the theory of roles developed. It exploits the power of task ontology and utilizes the task-specified roles and domain-specified roles.
    (3) A support system which can guide an author of an ontology has been built. The system is a product thanks to the theories and guideline we developed. It has been applied to a "real" problem, that is, construction of an oil-refinery plant ontology and its model. The evaluation shows that its performance is very satisfactory.
    The research thus can be concluded that it has made a contribution to the "content-oriented" research.

    researchmap

  • 複雑な知識構造を有す体系からの有意属性の構成的帰納基盤技術の研究

    Grant number:11878062  1999 - 2000

    日本学術振興会  科学研究費助成事業  萌芽的研究

    元田 浩, 堀内 匡, 鷲尾 隆

      More details

    Grant amount:\2100000 ( Direct Cost: \2100000 )

    研究実績は以下のとおり.
    1.混在属性に関する類似性尺度の検討
    数値属性,離散属性,名義属性のように性質の違った属性が混在した場合のデータの間の類似性を評価する手法を検討した.本来性質の違うもの同志の差を一般的に比較することは無理であるが,領域に依存した重み付けをすることによって単一の数値に写像する簡便な手法が実用的であることを示した.
    2.帰納的属性構成法の提案と評価
    複雑な構造体として与えられるデータは,基礎となる属性を組合わせて部分構造が形成され,これらがさらに再帰的に組み合わされて全体構造が形成されているものとみなすことができる.グラフに基づく帰納推論の手法は,一般グラフで表現される複雑な構造体に内在する部分構造を多頻度部分グラフとして発見することができる.この性質を帰納的属性構成に適用し,階層構造を有す部分構造を自動的に抽出し,新たな属性とする手法を提案した.その効果を変異源性の化合物の同定に適用し評価したところ,従来から知られている,親水性やエネルギー順位などの数値的な指標に劣らず,変異源性を同定するに有効な属性として機能することを確認した.
    3.帰納的属性構成法統合機械学習環境の構築
    上の結果を踏まえて前年度に設計した帰納的属性構成法を統合した機械学習環境を構築した.

    researchmap

  • Research on User-Adaptive Interface that Learns to Improve its Performance

    Grant number:09480065  1997 - 1999

    Japan Society for the Promotion of Science  Grants-in-Aid for Scientific Research  Grant-in-Aid for Scientific Research (B)

    MOTODA Hiroshi, HORIVCHI Tadashi, WASHIO Takashi, MIZOGVCH Riichiro

      More details

    Grant amount:\12300000 ( Direct Cost: \12300000 )

    Computer needs to come closer to humans if it is to be a good partner of its user as an easy-to-use tool by interpreting its user's intention, learning her preference and responding in accordance to her expertise of computer usage. This research targeted to develop a user-adaptive interface that learns user's preference from her past usage and predicts the next command. The key factor is to devise a right kind of machine learning technique that best suits to this needs. It was recognized that narrowing down the context in which the user was working is crucial, and to do this, use of the tree structured data that involve both command sequence data and process I/O data was essential. An efficient algorithm based on the notion of pairwise chunking was developed which enabled to induce a classifier in real time from a tree structured data. Evaluation results using both artificial and real data show that the prediction is accurate enough, and the implemented interface gradually improves its performance as it learns its user's preference and comes to respond differently for a different user. Further, the learning part was made an independent program and expanded to handle general graph structured data, i.e., directed/undirected graph that has colored/uncolored nodes and links with/out loops (including self-loops). Its computational complexity is confirmed to be linear to the size of graph. It was applied to finding typical patterns from WWW browsing history data provided by a commercial provider and to discovering characteristic substructures that are typical to carcinogen of organic chlorides. Both experiments gave satisfactory results.

    researchmap

  • 知識の世代交代が容易な可塑性型知識ベースの構築方法に関する研究

    Grant number:09878068  1997 - 1998

    日本学術振興会  科学研究費助成事業  萌芽的研究

    元田 浩, 堀内 匡, 鷲尾 隆

      More details

    Grant amount:\1900000 ( Direct Cost: \1900000 )

    本年度は,知識ベース構築理論の一層の改良に加えて,複数の専門家による知識獲得と大規模化に対する諸課題を検討した.
    1. 前年度に構築した理論を改良した.推論結果が求まらなかった場合に使用する暗黙解が,知識の獲得速度や知識ベースの大きさにどのような効果があるかを検討し,最適な暗黙解を選定するための規範を最小記述長原理に基づいて導出し,その効果を15種類の性質の異なるデータセットで検証した.提案した規範を用いることにより,知識更新が容易で,整合性の保持が保証された知識獲得性能のよいコンパクトな知識ベースが構築されることを確認した.
    2. 前年度に試作したプロトタイプシステムを改良した.データの属性値の種類を増やし,名辞属性の他に数値属性も扱えるようにした.簡単なユーザインターフェイスを加え,大規模化への対応と結果の視覚化を充実させた.
    3. 結論が複数ある場合への拡張や複数の専門家による独立した知識更新が可能となるような理論上の拡張を検討し,実用化システムの開発を目指した将来の研究課題をまとめた.

    researchmap

▼display all