Presentation 1996/12/14
Optimal Realization of Optimally Generalizing Neural Networks
Shinsuke NAKAZAWA, Hidemitsu OGAWA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Learning in a multilayered feedforward neural network is equivalent to finding an optimal function from the given set of training data, based on some optimization criterion. In the previous work, one of the authors gave a method of realizing a neural network which satisfies a criterion for generalization. In that realization, it was shown that there exists infinite freedom in the method of selecting the number of hidden units, basis functions, and weights parameters. In this paper, using the above mentioned degree of freedom, we obtain a realization of a neural network which is resistant to errors and faults in the hardware, while maintaining the generalization ablility.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) neural network / generalization / fault tolerance / connectlion fault / minimum error realization
Paper # NC96-60
Date of Issue

Conference Information
Committee NC
Conference Date 1996/12/14(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) Optimal Realization of Optimally Generalizing Neural Networks
Sub Title (in English)
Keyword(1) neural network
Keyword(2) generalization
Keyword(3) fault tolerance
Keyword(4) connectlion fault
Keyword(5) minimum error realization
1st Author's Name Shinsuke NAKAZAWA
1st Author's Affiliation Department of Computer Science Graduate School of Information Science and Engineering Tokyo Institute of Technology()
2nd Author's Name Hidemitsu OGAWA
2nd Author's Affiliation Department of Computer Science Graduate School of Information Science and Engineering Tokyo Institute of Technology
Date 1996/12/14
Paper # NC96-60
Volume (vol) vol.96
Number (no) 430
Page pp.pp.-
#Pages 8
Date of Issue