Presentation 1998/7/27
Discovery of Rules and Improvement of Generalization Power using SFNN and Destructive Structural Learning Algorithm
Shinichi Kikuchi, Masakazu Nakanishi,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we got a hint from Ishikawa's stuctural learning algorithm with frgetting to aim at discovery of rules and improvement of generalization power. A new point in order to discovery rules is to apply a new destructive structural learning algorithm to feed-forward neural networks based on layered structure(SFNN, Special Feed-forward Neural Networks). SFNN is a wider concept than a conventional multilayered stucture. Using this system, not only gazing spheres really appear on its structure but also stronger structures are attained in terms of its generalization power.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) neural networks / SFNN / structural learning / discovery of rules / generalization power
Paper # NC98-34
Date of Issue

Conference Information
Committee NC
Conference Date 1998/7/27(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) Discovery of Rules and Improvement of Generalization Power using SFNN and Destructive Structural Learning Algorithm
Sub Title (in English)
Keyword(1) neural networks
Keyword(2) SFNN
Keyword(3) structural learning
Keyword(4) discovery of rules
Keyword(5) generalization power
1st Author's Name Shinichi Kikuchi
1st Author's Affiliation Department of Science and Technology Graduate School of Keio University()
2nd Author's Name Masakazu Nakanishi
2nd Author's Affiliation Department of Science and Technology Graduate School of Keio University
Date 1998/7/27
Paper # NC98-34
Volume (vol) vol.98
Number (no) 219
Page pp.pp.-
#Pages 8
Date of Issue