Optimizing and Visualizing Kernel Classifiers

• Project/Area Number: 19360182
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: System engineering
• Research Institution: Kobe University
• Principal Investigator: ABE Shigeo Kobe University, 大学院工学研究科, 教授 (50294195)
• Co-Investigator(Kenkyū-buntansha): OZAWA Seiichi 神戸大学, 大学院工学研究科, 准教授 (70214129)
• Project Period (FY): 2007 – 2008
• Project Status: Completed (Fiscal Year 2008)
• Budget Amount: ¥7,020,000 (Direct Cost: ¥5,400,000、Indirect Cost: ¥1,620,000); Fiscal Year 2008: ¥3,120,000 (Direct Cost: ¥2,400,000、Indirect Cost: ¥720,000); Fiscal Year 2007: ¥3,900,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥900,000)
• Keywords: システム情報(知識)処理 / パターン認識 / サポートベクトルマシン / 特徴抽出 / 特徴選択 / ファジィ推論 / マルチプルクラシファイアシステム / カーネル判別分析 / オンライン特徴抽出

Research Abstract

a) KDA (Kernel Discriminant Analysis)を特徴選択の基準として特徴選択する方式を開発した.
b) 特徴空間上のKDA に基づいてパターン認識する方式を開発した.またファジィ識別器の可視化のプリミティブな方式を開発した.
c) カーネルファジィ識別器のメンバーシップ関数をSVM のマージン最大化の概念によりチューニングする方式を開発した.
d) 相関のある複数のパターン認識問題が逐次的に与えられるマルチタスク学習問題に対し,少ない訓練データで高い汎化能力が得られるマルチプルクラシファイアシステムを開発した.

Research Products

• [Journal Article] Decomposition Techniques for Training Linear Programming Support Vector Machines (2009)
  • Author(s): Yusuke Torii and Shigeo Abe
  • Journal Title: Neurocomputing 72 Pages: 973-984
  • NAID: 120001478926
  • Peer Reviewed
• [Journal Article] A Multitask Learning Model for Online Pattern Recognition (2009)
  • Author(s): Seiichi Ozawa, Asim Roy, and Dmitri Roussinov
  • Journal Title: IEEE Trans. on Neural Networks 20 Pages: 430-445
  • NAID: 120001609316
  • Peer Reviewed
• [Journal Article] Decomposition Techniques for Training Linear Programming Support Vector Machines (2009)
  • Author(s): Yusukc Torii
  • Journal Title: Neurocompuling 72 Pages: 973-984
  • Peer Reviewed
• [Journal Article] Kernel Discriminant Analysis Based Feature Selection (2008)
  • Author(s): Tsuneyoshi Ishii, Masamichi Ashihara, and Shigeo Abe
  • Journal Title: Neurocomputing 71 Pages: 2544-2552
  • NAID: 120001478927
  • Peer Reviewed
• [Journal Article] Incremental Learning of Chunk Data for On-line Pattern Classification Systems (2008)
  • Author(s): Seiichi Ozawa, Shaoning Pang, and Nikola Kasabov
  • Journal Title: IEEE Trans. on Neural Networks 19 Pages: 1061-1074
  • Peer Reviewed
• [Journal Article] Kernel Discriminant Analysis Based Feature Selection (2008)
  • Author(s): Tsuneyoshi Ishii
  • Journal Title: Neurocomputing 71 Pages: 1853-1863
  • Peer Reviewed
• [Journal Article] Comparison of sparse least squares support vector regressors trained in primal and dual (2008)
  • Author(s): Shigco Abe
  • Journal Title: Proc. ESANN 2008 Pages: 469-474
  • Peer Reviewed
• [Journal Article] Sparse Support Veclor Machines Trained in the Reduced Empirical Feature Space (2008)
  • Author(s): Kazuki Iwamura
  • Journal Title: Proc. IJCNN 2008 Pages: 2399-2405
  • Peer Reviewed
• [Journal Article] Feature Selection Based on Kernel Discriminant Anialysis for Multi-class Problems (2008)
  • Author(s): Tsunekichi Ishii
  • Journal Title: Proc. IJCNN 2008 Pages: 2456-2461
  • Peer Reviewed
• [Journal Article] sparse Least Squares Support Vector Machines by Forward Selection Based on Linear Discriminanl Analysis (2008)
  • Author(s): Shigeo Abe
  • Journal Title: Proc. ANNPR 2008 Pages: 54-65
  • Peer Reviewed
• [Journal Article] Batcch Support Vector Training Based on Exact Incremental Training (2008)
  • Author(s): Shigeo Abe
  • Journal Title: Proc. ICANN 2008 Pages: 295-304
  • Peer Reviewed
• [Journal Article] Incremental Learning of Chunk Data for On-line Pattern Classification Systems (2008)
  • Author(s): Seiichi Ozawa
  • Journal Title: IEEE Trans. on Neural Networks 19 Pages: 1061-1074
  • Peer Reviewed
• [Journal Article] Incremental Learning for Multitask Pattern Recognition Problems (2008)
  • Author(s): Seiichi Ozawa
  • Journal Title: Proc. ICMLA2008 Pages: 747-751
  • Peer Reviewed
• [Journal Article] Adaptive Face Recognition System Using Fast Incremental Principal Component Analysis (2008)
  • Author(s): Seiichi Ozawa
  • Journal Title: Proc. of 14th Int. Conf. on Neural Information Processing (印刷中)
  • Peer Reviewed
• [Journal Article] A Learning Algorithm of Boosting Kernel Discriminant Analysis for Pattern Recognition (2007)
  • Author(s): Shinji Kita, Seiichi Ozawa, Satoshi Maekawa, and Shigeo Abe
  • Journal Title: IEICE Trans. Information Systems E90-D Pages: 1853-1863
  • NAID: 110007541161
  • Peer Reviewed
• [Journal Article] Sparse Least Squares Support Vector Training in the Reduced Empirical Feature Space (2007)
  • Author(s): Shigeo Abe
  • Journal Title: Pattern Analysis & Applications 10 Pages: 203-214
  • NAID: 120000945576
  • Peer Reviewed
• [Journal Article] Backward Variable Selection of Support Vector Regressors by Block Deletion (2007)
  • Author(s): Takashi Nagatani
  • Journal Title: Proc. International Joint Conference on Neural Networks Pages: 1540-1545
  • Peer Reviewed
• [Journal Article] Fuzzy Classifiers Based on Kernel Discriminant Analysis (2007)
  • Author(s): Ryota Hosokawa
  • Journal Title: International Conference on Artificial Neural Networks(ICANN 2007) Part II Pages: 180-189
  • Peer Reviewed
• [Journal Article] A Learning Algorithm of Boosting Kernel Discriminant Analysis for Pattern Classification (2007)
  • Author(s): Shinji Kita
  • Journal Title: IEICE Trans. on Information and Systems E90-D Pages: 1853-1863
  • Peer Reviewed
• [Journal Article] Boosting Kernel Discriminant Analysis for Pattern Classification (2007)
  • Author(s): Shinji Kita
  • Journal Title: Proc. of Int. Symposium on Intelligent Signal Processing and Communication Systems Pages: 854-857
  • Peer Reviewed
• [Journal Article] An Efficient Incremental Kernel Principal Component Analysis for Online Feature Selection (2007)
  • Author(s): Yohei Takeuchi
  • Journal Title: Proc. Int. Joint Conf. on Neural Networks 2007 (CD-ROM)
  • Peer Reviewed
• [Journal Article] A Multitask Learning Model for Online Pattern Recognition
  • Author(s): Seiichi Ozawa
  • Journal Title: IEEE Trans. on Neural Networks (印刷中)(掲載確定)
  • NAID: 120001609316
  • Peer Reviewed
• [Journal Article] Incremental Principal Component Analysis Based on Adaptive Accumulation Ratio
  • Author(s): Seiichi Ozawa
  • Journal Title: Proc. ICONIP2008 (印刷中)(掲載確定)
  • Peer Reviewed
• [Journal Article] A Novel Incremental Linear Discriminant Analysis for Multitask Pattern Recognition Problems
  • Author(s): Masayuki Hisada
  • Journal Title: Proc. ICONIP2008 (印刷中)(掲載確定)
  • Peer Reviewed
• [Journal Article] A Neural Network Model for Sequential Multitask Pattern Recognition Problems
  • Author(s): Hitoshi Nishikawa
  • Journal Title: Proc. ICONIP2008 (印刷中)(掲載確定)
  • Peer Reviewed
• [Journal Article] Kernel Discriminant Analysis Based Feature Selection
  • Author(s): Tsuneyoshi Ishii
  • Journal Title: Neurocomputing (印刷中)
  • Peer Reviewed
• [Journal Article] Feature Selection Based on Kernel Discriminant Analysis for Multi-Class Problems
  • Author(s): Tsuneyoshi Ishii
  • Journal Title: Proc. International Joint Conference on Neural Networks(IJCNN 2008) (印刷中)
  • Peer Reviewed
• [Presentation] An Incremental Principal Component Analysis Based on Dynamic Accumulation Eatio (2008)
  • Author(s): Seiichi Ozawa
  • Organizer: Proc. of SICE Annual Conf.
  • Place of Presentation: 山気通信大学
  • Year and Date: 2008-08-21
• [Presentation] 追加学習型主成分分析の改良 (2008)
  • Author(s): 松本和也
  • Organizer: システム制御悄報学会研究発表m演会
  • Place of Presentation: 京都情報大学院大学
  • Year and Date: 2008-05-17
• [Presentation] Incremental Learning for Multitask Pattern Recognition Problems, (2008)
  • Author(s): Seiichi Ozawa and Asim Roy
  • Organizer: Proc.ICMLA 2008
  • Place of Presentation: San Diego, USA
• [Presentation] ICONIP 2008 (in press) (2008)
  • Author(s): Masayuki Hisada, Seiichi Ozawa, Kau Zhang, Shaoning Pang, and Nikola Kasabov
  • Organizer: A Novel Incremental Linear Discriminant Analysis for Multitask Pattern Recognition Problems, Proc.
  • Place of Presentation: Auckland, New Zealand
• [Presentation] ICONIP 2008 (in press) (2008)
  • Author(s): Hitoshi Nishikawa, Seiichi Ozawa, and Asim Roy
  • Organizer: A Neural Network Model for Sequential Multitask Pattern Recognition Problems, Proc.
  • Place of Presentation: Auckland, New Zealand
• [Presentation] ICONIP 2008 (in press) (2008)
  • Author(s): Seiichi Ozawa, Kazuya Matsumoto, Shaoning Pang, and Nikola Kasabov
  • Organizer: Incremental Principal Component Analysis Based on Adaptive Accumulation Ratio, Proc.
  • Place of Presentation: Auckland, New Zealand
• [Presentation] ANNPR 2008 (2008)
  • Author(s): Shigeo Abe
  • Organizer: Sparse Least Squares Support Vector Machines by Forward Selection Based on Linear Discriminant Analysis, Proc.
  • Place of Presentation: Paris, France
• [Presentation] IJCNN (2008)
  • Author(s): Tsunekichi Ishii and Shigeo Abe
  • Organizer: Feature Selection Based on Kernel Discriminant Analysis for Multi-class Problems, Proc.
  • Place of Presentation: Hong Kong, China
• [Presentation] IJCNN 2008 (2008)
  • Author(s): Kazuki Iwamura and Shigeo Abe
  • Organizer: Sparse Support Vector Machines Trained in the Reduced Empirical Feature Space, Proc.
  • Place of Presentation: Hong Kong, China
• [Presentation] 特徴選択による追加学習型カーネル主成分分析の高速化とその性能評価 (2007)
  • Author(s): 竹内洋平
  • Organizer: 平成19年電気関係学会関西支部連合大会
  • Place of Presentation: 神戸大学
  • Year and Date: 2007-11-18
• [Presentation] An Online Face Recognition System with Incremental Learning Ability (2007)
  • Author(s): Seiichi Ozawa
  • Organizer: Proc. of SICE Annual Conf. 2007
  • Place of Presentation: 香川大学
  • Year and Date: 2007-09-20
• [Presentation] 追加学習型カーネル主成分分析の開発とその性能評価 (2007)
  • Author(s): 竹内洋平
  • Organizer: 平成19年電気学会電子・情報・システム部門大会
  • Place of Presentation: 大阪府立大学
  • Year and Date: 2007-09-04
• [Presentation] ISPACS 2007 (2007)
  • Author(s): Shinji Kita, Seiichi Ozawa, Satoshi Maekawa, and Shigeo Abe
  • Organizer: Boosting Kernel Discriminant Analysis for Pattern Classification, Proc.
  • Place of Presentation: Xiamen, China
• [Presentation] Lecture Notes in Computer Science 4985, Springer (2007)
  • Author(s): Seiichi Ozawa, Shaoning Pang, and Nikola Kasabov
  • Organizer: Adaptive Face Recognition System Using Fast Incremental Principal Component Analysis, in Neural Information Processing
  • Place of Presentation: Kitakyushu, Japan
• [Presentation] ICANN 2007 (2007)
  • Author(s): Ryota Hosokawa and Shigeo Abe
  • Organizer: Fuzzy Classifiers Based on Kernel Discriminant Analysis, Proc.
  • Place of Presentation: Porto, Portugal
• [Presentation] IJCNN 2007 (2007)
  • Author(s): Takashi Nagatani and Shigeo Abe
  • Organizer: Backward Variable Selection of Support Vector Regressors by Block Deletion, Proc.
  • Place of Presentation: Orlando, Florida, USA
• [Presentation] IJCNN 2007 (2007)
  • Author(s): Yohei Takeuchi, Seiichi Ozawa, and Shigeo Abe
  • Organizer: An Efficient Incremental Kernel Principal Component Analysis for Online Feature Selection, Proc.
  • Place of Presentation: Orlando, Florida, USA
• [Book] Online Incremental Face Recognition System Using Eigenface Feature and Neural Classifier, in State of the Art in Face Recognition, Julio Ponce and Adem Karahoca, Eds. (2009)
  • Author(s): Seiichi Ozawa, Shigeo AbeAbe, Shaoning Pang, and Nikola Kasabov
  • Publisher: IN-TECH
• [Book] Optimizing Mahalanobis Kernels for Pattern Classification, in Pattern Recognition: Theory and Application, E. A. Zoeller, Ed. (2008)
  • Author(s): Shigeo Abe
  • Publisher: Nova Science Publishers