Presentation 2010-03-11
A Calculation of Three Layer Neural Network Structures Approximating Polynomial Functions
Yuichi NAKAMURA, Masahiro NAKAGAWA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The function approximation theorem has been proved such that the suitably constructed feedforward neural networks can sufficiently approximate given continuous functions. However, the specific structures of feed- forward networks approximating functions has not been treated in the function approximation theorem. In this report, a specific structure of three layer feedforward networks approximating polynomial functions is derived. The feedforward network has some system parameters. The structure of the feedforward network is determined by the parameters. The system parameters are systematically derived by the Taylor expansion of the activation function and by the matrix operation when the activation function is smooth. It can be expected that the three layer feedforward network structure obtained from the proposed method gives the effective information to learning algorithms of the neural networks.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Feed-forward neural network / Structure calculation / Function approximation / Polynomial
Paper # NC2009-167
Date of Issue

Conference Information
Committee NC
Conference Date 2010/3/2(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) A Calculation of Three Layer Neural Network Structures Approximating Polynomial Functions
Sub Title (in English)
Keyword(1) Feed-forward neural network
Keyword(2) Structure calculation
Keyword(3) Function approximation
Keyword(4) Polynomial
1st Author's Name Yuichi NAKAMURA
1st Author's Affiliation Nagaoka University of Technology()
2nd Author's Name Masahiro NAKAGAWA
2nd Author's Affiliation Nagaoka University of Technology
Date 2010-03-11
Paper # NC2009-167
Volume (vol) vol.109
Number (no) 461
Page pp.pp.-
#Pages 6
Date of Issue