A Study on nonlinear classification methods based on the optimization of the structure of the reproducing kernel Hilbert space

• Project/Area Number: 25870811
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Intelligent informatics; Statistical science
• Research Institution: University of Tsukuba
• Principal Investigator: HINO Hideitsu 筑波大学, システム情報系, 助教 (10580079)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Project Status: Completed (Fiscal Year 2015)
• Budget Amount: ¥3,380,000 (Direct Cost: ¥2,600,000、Indirect Cost: ¥780,000); Fiscal Year 2015: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000); Fiscal Year 2014: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000); Fiscal Year 2013: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
• Keywords: 機械学習 / 応用統計 / パターン認識 / カーネル法 / 類似度 / エントロピー / 情報量 / 多変量解析

Outline of Final Research Achievements

Kernel method is one of the most important methods in machine learning. Its effectiveness depends on the kernel function and its parameter. We developed methods for optimizing the kernel function based on the data distribution of the intrinsic high dimensional space associated with the kernel function. We also proposed a similarity measure for the given data and the newly observed data based on the notion of information content in the given data, and applied the proposed measure for classifying a set of data and finding outliers from the observed dataset. We applied the developed methods to speaker recognition, hand gesture recognition, protein structure classification, and change point detection from time series data.

Research Products

• [Journal Article] Protein Fold Classification Using Large Margin Combination of Distance Metrics (2016)
  • Author(s): Chendra Hadi Suryanto, Kazuhiro Fukui, Hideitsu Hino
  • Journal Title: IEICE Transactions on Information and Systems Volume: E99.D Issue: 3 Pages: 714-723
  • DOI: 10.1587/transinf.2015EDP7294
  • NAID: 130005131822
  • ISSN: 0916-8532, 1745-1361
  • Peer Reviewed / Acknowledgement Compliant
• [Journal Article] mmpp: a Package for Calculating Similarity and Distance Metrics for Simple and Marked Temporal Point Process (2015)
  • Author(s): Hideitsu Hino, Ken Takano, Noboru Murata
  • Journal Title: The R Journal Volume: 7
  • Peer Reviewed / Open Access / Acknowledgement Compliant
• [Journal Article] Change-Point Detection in a Sequence of Bags-of-Data (2015)
  • Author(s): Kensuke Koshijima, Hideitsu Hino, Noboru Murata
  • Journal Title: IEEE Transactions on Knowledge and Data Engineering Volume: 27 Issue: 10 Pages: 2632-2644
  • DOI: 10.1109/tkde.2015.2426693
  • NAID: 120007136176
  • Peer Reviewed / Acknowledgement Compliant
• [Journal Article] Non-Parametric Entropy Estimators Based on Simple Linear Regression (2015)
  • Author(s): Hideitsu Hino, Kensuke Koshijima, Noboru Murata
  • Journal Title: Computational Statistics and Data Analysis Volume: 89 Pages: 72-84
  • DOI: 10.1016/j.csda.2015.03.011
  • Peer Reviewed / Acknowledgement Compliant
• [Journal Article] Entropy Power Inequality for Learning Optimal Combination of Kernel Functions (2015)
  • Author(s): Hideitsu Hino
  • Journal Title: Journal of Signal Processing Systems Volume: 79(2) Issue: 2 Pages: 201-210
  • DOI: 10.1007/s11265-014-0899-7
  • Peer Reviewed / Acknowledgement Compliant
• [Journal Article] Intrinsic Graph Structure Estimation Using Graph Laplacian (2014)
  • Author(s): A. Noda, H. Hino, M. Tatsuno, S. Akaho, N. Murata
  • Journal Title: Neural Computation Volume: 26 Issue: 7 Pages: 1455-1483
  • DOI: 10.1162/neco_a_00603
  • Peer Reviewed / Acknowledgement Compliant
• [Journal Article] A Non-parametric Clustering Algorithm With A Quantile-based Likelihood Estimator (2014)
  • Author(s): Hideitsu Hino, Noboru Murata
  • Journal Title: Neural Computation Volume: 26(9) Issue: 9 Pages: 2074-2101
  • DOI: 10.1162/neco_a_00628
  • Peer Reviewed / Acknowledgement Compliant
• [Journal Article] A kernel method to extract common features based on mutual information (2014)
  • Author(s): T. Araki, H. Hino, S. Akaho
  • Journal Title: Lecture Notes in Computer Science (Neural Information Processing) Volume: 8835 Pages: 26-34
  • DOI: 10.1007/978-3-319-12640-1_4
  • ISBN: 9783319126395, 9783319126401
  • Peer Reviewed / Open Access / Acknowledgement Compliant
• [Journal Article] A Non-Parametric Maximum Entropy Clustering (2014)
  • Author(s): Hideitsu Hino, Noboru Murata
  • Journal Title: 24th International Conference on Artificial Neural Networks, Hamburg, Germany, September 15-19, 2014. Proceedings Volume: 8681 Pages: 113-120
  • DOI: 10.1007/978-3-319-11179-7_15
  • ISBN: 9783319111780, 9783319111797
  • Peer Reviewed
• [Journal Article] An Algorithm for Directed Graph Structure Estimation (2014)
  • Author(s): H. Hino, A. Noda, M. Tatsuno, S. Akaho, N. Murata
  • Journal Title: Lecture Notes in Computer Science (Artificial Neural Networks and Machine Learning) Volume: 8681 Pages: 145-152
  • DOI: 10.1007/978-3-319-11179-7_19
  • ISBN: 9783319111780, 9783319111797
  • Peer Reviewed / Acknowledgement Compliant
• [Journal Article] Information estimators for weighted observations (2013)
  • Author(s): Hideitsu Hino, Noboru Murata
  • Journal Title: Neural Networks Volume: 46 Pages: 260-275
  • DOI: 10.1016/j.neunet.2013.06.005
  • NAID: 120007136808
  • Peer Reviewed
• [Journal Article] Pairwise Similarity for Line Extraction from Distorted Images (2013)
  • Author(s): Hideitsu Hino, Jun Fujiki, Shotaro Akaho, Yoshihiko Mochizuki, Noboru Murata
  • Journal Title: Lecture Notes in Computer Science Volume: 8048 Pages: 250-257
  • DOI: 10.1007/978-3-642-40246-3_31
  • ISBN: 9783642402456, 9783642402463
  • Peer Reviewed
• [Journal Article] Gaussian Multiple Kernel Learning with Entropy Power Inequality (2013)
  • Author(s): Hideitsu Hino
  • Journal Title: In Proceedings of Machine Learning for Signal Processing (MLSP), 2013 IEEE International Workshop on Volume: 1 Pages: 1-6
  • DOI: 10.1109/mlsp.2013.6661956
  • Peer Reviewed
• [Journal Article] Supervised Covariance Selection For Linear Discriminant Analysis (2013)
  • Author(s): Hideitsu Hino, Nima Reyhani
  • Journal Title: In Proceedings of Machine Learning for Signal Processing (MLSP), 2013 IEEE International Workshop on Volume: 1 Pages: 1-6
  • DOI: 10.1109/mlsp.2013.6661942
  • Peer Reviewed
• [Journal Article] Integration of MKL-based and i-vector-based speaker verification by short utterances (2013)
  • Author(s): Hideitsu Hino, Tetsuji Ogawa
  • Journal Title: In Proceedings of Pattern Recognition (ACPR), 2013 2nd IAPR Asian Conference on Volume: 1 Pages: 562-566
  • DOI: 10.1109/acpr.2013.42
  • Peer Reviewed
• [Journal Article] Combination of Multiple Distance Measures for Protein Fold Classification (2013)
  • Author(s): Chendra Hadi Suryanto, Hideitsu Hino, Kazuhiro Fukui
  • Journal Title: In Proceedings of Pattern Recognition (ACPR), 2013 2nd IAPR Asian Conference on Volume: 1 Pages: 440-445
  • DOI: 10.1109/acpr.2013.139
  • Peer Reviewed
• [Presentation] Spotting Finger Spelled Words from Sign Language Video by Temporally Regularized Canonical Component Analysis (2016)
  • Author(s): Shohei Tanaka, Hideitsu Hino, Kazuhiro Fukui
  • Organizer: The IEEE International Conference on Identity, Security and Behavior Analysis 2016
  • Place of Presentation: Tohoku University, Miyagi, Japan
  • Year and Date: 2016-02-29
  • Int'l Joint Research
• [Presentation] Personal Authentication based on 3D Configuration of Micro-Feature Points on Facial Surface (2015)
  • Author(s): Takao Yoshinuma, Hideitsu Hino and Kazuhiro Fukui
  • Organizer: The 7the Pacific-Rim Symposium on Image and Video Technology
  • Place of Presentation: Auckland, New Zealand
  • Year and Date: 2015-11-23
• [Presentation] 機械学習の考え方～データからの知識発見に向けて～ (2015)
  • Author(s): 日野英逸
  • Organizer: 2015年電子情報通信学会 総合大会
  • Place of Presentation: 立命館大学，滋賀
  • Year and Date: 2015-03-10 – 2015-03-13
  • Invited
• [Presentation] 単回帰に基づく微分エントロピー推定量 (2015)
  • Author(s): 日野英逸
  • Organizer: 大規模統計モデリングと計算統計研究会
  • Place of Presentation: 東京大学，東京
  • Year and Date: 2015-02-06 – 2015-02-07
• [Presentation] 確率質量関数の二次展開と単回帰に基づくエントロピー推定 (2014)
  • Author(s): 日野英逸
  • Organizer: 情報論的学習理論と機械学習研究会
  • Place of Presentation: 名古屋大学，愛知
  • Year and Date: 2014-11-17 – 2014-11-19
• [Presentation] A Non-Parametric Maximum Entropy Clustering (2014)
  • Author(s): Hideitsu Hino
  • Organizer: International Conference on Artificial Neural Networks 2014
  • Place of Presentation: Hamburg, Germany
  • Year and Date: 2014-09-15 – 2014-09-19
• [Presentation] An Algorithm for Directed Graph Estimation (2014)
  • Author(s): Hideitsu Hino
  • Organizer: International Conference on Artificial Neural Networks 2014
  • Place of Presentation: Hamburg, Germany
  • Year and Date: 2014-09-15 – 2014-09-19
• [Presentation] A kernel method to extract common features based on mutual information (2014)
  • Author(s): Takamitsu Araki
  • Organizer: The 21st International Conference on Neural Information Processing
  • Place of Presentation: Kuching, Malaysia
  • Year and Date: 2014-09-03 – 2014-09-06
• [Presentation] スパースモデリングの基礎（チュートリアル講演） (2014)
  • Author(s): 日野英逸
  • Organizer: 第17回画像の認識・理解シンポジウム
  • Place of Presentation: 岡山コンベンションセンター, 岡山
  • Year and Date: 2014-07-28 – 2014-07-31
  • Invited
• [Presentation] Information and entropy estimators for machine learning and pattern recognition (2014)
  • Author(s): Hideitsu Hino
  • Organizer: The 3nd Institute of Mathematical Statistics Asia Pacific Rim Meeting
  • Place of Presentation: Taipei, Taiwan
  • Year and Date: 2014-06-29 – 2014-07-03
• [Presentation] 重み付き有向グラフモデリングによるスパイクデータ解析 (2014)
  • Author(s): 樋口翔
  • Organizer: ニューロコンピューティング研究会
  • Place of Presentation: 沖縄科学技術大学院大学，沖縄
  • Year and Date: 2014-06-25 – 2014-06-27
• [Presentation] データ対の直線性に基づく画像上の類似度の定義 ～ 歪曲画像からの直線検出への応用 ～ (2013)
  • Author(s): 日野英逸, 藤木淳，赤穂昭太郎，望月義彦，村田昇
  • Organizer: パターン認識・メディア理解研究会
  • Place of Presentation: NTT武蔵野研究開発センタ