Fast Training of Three-Layered Neural Network Classifiers by Solving Inequalities

• TSUCHIYA Naoki

  Graduate School of Science and Technology, Kobe University

• OZAWA Seiichi

  Graduate School of Science and Technology, Kobe University

• ABE Shigeo

  Graduate School of Science and Technology, Kobe University

Bibliographic Information

Other Title

• パターン認識用三層ニューラルネットの不等式の求解による高速学習方式
• パターン ニンシキヨウ 3ソウ ニューラルネット ノ フトウシキ ノ キュウカイ ニ ヨル コウソク ガクシュウ ホウシキ

Abstract

In this paper we discuss training of three-layered neural network classifiers by solving inequalities. Namely, first we represent each class by the center of the training data belonging to the class, and determine the set of hyperplanes that separate each class (i.e., each center) into a single region. Then according to whether the center is on the positive or negative side of the hyperplane, we determine the target values of each class for the hidden neurons (i.e., hyperplanes). Since the convergence condition of the neural network classifier is now represented by the two sets of inequalities, we solve the sets successively by the Ho-Kashyap algorithm. We demonstrate the advantage of our method over the backpropagation algorithm using several benchmark data sets.

Journal

• Transactions of the Institute of Systems, Control and Information Engineers

  Transactions of the Institute of Systems, Control and Information Engineers 13 (6), 276-283, 2000

  THE INSTITUTE OF SYSTEMS, CONTROL AND INFORMATION ENGINEERS (ISCIE)

Keywords

• backpropagation algorithm
• classifiers
• HoKashyap algorithm
• inequalities
• neural networks

Details

• CRID: 1390001205165188608
• NII Article ID: 120001642383
• NII Article ID: 10004473143
• NII Book ID: AN1013280X
• DOI: 10.5687/iscie.13.6_276
• ISSN: 2185811X
• ISSN: 13425668
• HANDLE: 20.500.14094/90001021
• NDL BIB ID: 5370188
• Web Site: https://ndlsearch.ndl.go.jp/books/R000000004-I5370188
• Data Source:
  • JaLC
  • IRDB
  • NDL
  • Crossref
  • CiNii Articles
• Abstract License Flag: Disallowed