Title:
【発明の名称】ニューラルネットワークにおける結合のグループ単位学習調整方式
Document Type and Number:
Japanese Patent JP3172163
Kind Code:
B2
Abstract:
PURPOSE:To convert the construction of the neural network and to perform the fine adjustment of the degree of the learning of respective connections by setting the output of a learning adjustment constant output means to a learning adjustment device for the required connection through a neural network learning means and controlling the learning for the unit of a group. CONSTITUTION:A learning adjustment constant output means 3 outputs the required learning adjustment constant by receiving the instruction to set plural groups composed of the same phase of the unit composing the neural network or composed of plural units spreading over plural phases by the reverse propagation of the learning adjustment constant setting to learning adjustors 2a-2n to respective connections 1a-1n of the inside of the more than one group among the set groups. These constants are set through a neural network learning means 4 to the designated devices 2a-2n to control the neural network learning for the unit of the group so that the construction of the neural network can be converted to certain specific construction and the fine adjustment of the learning of respective connections can be performed.
Inventors:
Ryusuke Masuoka
Nobuo Watanabe
Asahi Kawamura
Yuri Owada
Kazuo Asakawa
Shigenori Matsuoka
Hiroyuki Okada
Nobuo Watanabe
Asahi Kawamura
Yuri Owada
Kazuo Asakawa
Shigenori Matsuoka
Hiroyuki Okada
Application Number:
JP11578790A
Publication Date:
June 04, 2001
Filing Date:
May 07, 1990
Export Citation:
Assignee:
富士通株式会社
株式会社エフ・エフ・シー
株式会社エフ・エフ・シー
International Classes:
G06G7/60; G06F15/18; G06N3/08; G06N99/00; (IPC1-7): G06G7/60; G06N3/08
Other References:
【文献】情報処理学会第40回(平成2年前期)全国大会講演論文集3C-10 渡部信雄、川村旭他「ニューラルネットワークによるファジィ制御-ニューロ・ファジィ融合システムの検討-」
Attorney, Agent or Firm:
Yoshiyuki Osuge (1 outside)