Presentation 2005-12-09
Non-linear gating network for the large scale classification model CombNET-II
Mauricio KUGLER, Toshiyuki MIYATANI, Susumu KUROYANAGI, Akira IWATA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The linear gating classifier (stem network) of the large scale model CombNET-II had been always the limitation factor for increasing the number of expert classifiers (branch networks). The linear boundaries between its clusters cause a rapidly decrease of the performance with the increase of the number of clusters and, consequently, impairing the whole structure performance. This work proposes the use of a non-linear classifier to learn the complex boundaries between the clusters, increasing the gating performance while keeping the balanced split of samples produced by the original sequential clustering algorithm. The experiments showed that, for some problems, the proposed model overcomes even the monolithic classifier.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) large scale classification problems / divide-and-conquer / gating network / sequential clustering
Paper # NC2005-87
Date of Issue

Conference Information
Committee NC
Conference Date 2005/12/2(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 ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Non-linear gating network for the large scale classification model CombNET-II
Sub Title (in English)
Keyword(1) large scale classification problems
Keyword(2) divide-and-conquer
Keyword(3) gating network
Keyword(4) sequential clustering
1st Author's Name Mauricio KUGLER
1st Author's Affiliation The authors are with the Department of Computer Science & Engineering, Nagoya Institute of Technology()
2nd Author's Name Toshiyuki MIYATANI
2nd Author's Affiliation The authors are with the Department of Computer Science & Engineering, Nagoya Institute of Technology
3rd Author's Name Susumu KUROYANAGI
3rd Author's Affiliation The authors are with the Department of Computer Science & Engineering, Nagoya Institute of Technology
4th Author's Name Akira IWATA
4th Author's Affiliation The authors are with the Department of Computer Science & Engineering, Nagoya Institute of Technology
Date 2005-12-09
Paper # NC2005-87
Volume (vol) vol.105
Number (no) 457
Page pp.pp.-
#Pages 6
Date of Issue