Presentation 1996/2/3
Direct inverse model learning by fluctuation-driven learning
Hirotada FURUKAWA, Takahumi OOHORI, Kazuhisa WATANABE,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) It is not simple for a neural network(NN) to learn the inverse model of unknown target system by means of conventional back-propagation(BP) rule. The feedback system to determine teacher's signals makes the total system complicate and of large scale. We show that NN can learn the inverse model comparably easily by means of the fluctuation-driven learning(FDL) rule. The FDL rule doesn't need neither differential process nor back-propagating process. Therefore, the feedback system is extremely simple and NN can consist of both continuous and discrete neurons. Simulations for inverse models of polinomial functions demonstrate the validity of our method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) inverse model / fluctuation-driven learning rule
Paper # NC95-109
Date of Issue

Conference Information
Committee NC
Conference Date 1996/2/3(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Direct inverse model learning by fluctuation-driven learning
Sub Title (in English)
Keyword(1) inverse model
Keyword(2) fluctuation-driven learning rule
1st Author's Name Hirotada FURUKAWA
1st Author's Affiliation Department of Electrical Engineering, Hokkaido Institute of Technology()
2nd Author's Name Takahumi OOHORI
2nd Author's Affiliation Department of Electrical Engineering, Hokkaido Institute of Technology
3rd Author's Name Kazuhisa WATANABE
3rd Author's Affiliation Department of Electrical Engineering, Hokkaido Institute of Technology
Date 1996/2/3
Paper # NC95-109
Volume (vol) vol.95
Number (no) 506
Page pp.pp.-
#Pages 8
Date of Issue