Presentation 2004/10/11
Dynamic Construction of Fault Tolerant Multi-layer Neural Networks
Ayumi NOBUTO, Haruhiko TAKASE, Hidehiko KITA, Terumine HAYASHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we aim to enhance the fault tolerance of multi-layer neural networks and to construct them with only essential units. In general, the procedure of constructing networks is as follows: (1) deciding the network structure (including the number of units) (2) training the network. To find desired network, it is necessary to repeat this procedure. To avoid the repetition, dynamic constructive algorithms are proposed. They change the network structure with the progress of the training. We combine a dynamic constructive algorithm with the training algorithm that enhance fault tolerance for networks whose structure is fixed.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Multi-Layer Neural Network / Fault Tolerant / Dynamic Construction
Paper # NC2004-58
Date of Issue

Conference Information
Committee NC
Conference Date 2004/10/11(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) Dynamic Construction of Fault Tolerant Multi-layer Neural Networks
Sub Title (in English)
Keyword(1) Multi-Layer Neural Network
Keyword(2) Fault Tolerant
Keyword(3) Dynamic Construction
1st Author's Name Ayumi NOBUTO
1st Author's Affiliation Faculty of Engineering, Mie University()
2nd Author's Name Haruhiko TAKASE
2nd Author's Affiliation Faculty of Engineering, Mie University
3rd Author's Name Hidehiko KITA
3rd Author's Affiliation Faculty of Engineering, Mie University
4th Author's Name Terumine HAYASHI
4th Author's Affiliation Faculty of Engineering, Mie University
Date 2004/10/11
Paper # NC2004-58
Volume (vol) vol.104
Number (no) 348
Page pp.pp.-
#Pages 6
Date of Issue