Back

Title
  • 情報量最大化学習を用いたニューラルネットワークの特性
Alternative
  • Features of Neural Networks Using Learning Algorithm Based on Maximizing Amount of Information
Creator

姉川, 正紀

志久, 修

中村, 千秋

中村, 彰

Contributor
長崎大学工学部電気情報工学科
佐世保工業高等専門学校電子制御工学科
長崎大学工学部電気情報工学科
長崎大学工学部電気情報工学科
Department of Electrical Engineering and Computer Science, The Faculty of Engineering, Nagasaki University
Department of Control Engineering, Sasebo Collage
Department of Electrical Engineering and Computer Science, The Faculty of Engineering, Nagasaki University
Department of Electrical Engineering and Computer Science, The Faculty of Engineering, Nagasaki University
Description
Other
  • A new learning algorithm based on the amount of information is proposed in this paper. A neuron pool (a group of neurons) is adopted as a basic information processing unit and the neuron-pools-network is self organized according to the proposed rule invented imitating 'Homeostasis' (maintain the constancy of life) of a living body. Learning is carried out in such a way that the amount of information of each neuron pool brings to maximum. Simulation was carried out using the simple neuron-pools-networks that neuron pools are alined in one dimension and multiple layers. The results show that it behaves similar to a living body and has pattern recognition ability to some extent.
Other
  • identifier:長崎大学工学部研究報告 Vol.26(46) p. 31-37, 1996
Date Issued 1996-01
Languagejpn
NIItypedepartmental bulletin paper
VersiontypeNA
Identifier URI http://hdl.handle.net/10069/14963
Journal
    • ISSN 0286-0902
    • 長崎大学工学部研究報告
    26(46), 31-37
File
Oaidate2012-01-26T06:47:37Z