Presentation | 1997/10/20 Learning Processes of Neural Networks and Error Meaures Sumiyoshi Fujiki, M. Nahomi Fujiki, Mitsuyuki Nakao, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In the learning process by the conventional back-propagation algorithm, the trapping problem in metastable states occurs using the quadratic error measure. By using the Kullback divergence as an error measure, it is shown by numerical simulations that the learning efficiency is improved significantly no matter how large the number of internal nurons, and the optimal network size is achieved automatically. The Kullback divergence is a superior error measure in the two sence of the scalability: one for the complexity of the problem, and another for the excess network size. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Layered Neural Network / Learning Process / Error back-propagation / Kullback's Measure |
Paper # | NC97-43 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 1997/10/20(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) | Learning Processes of Neural Networks and Error Meaures |
Sub Title (in English) | |
Keyword(1) | Layered Neural Network |
Keyword(2) | Learning Process |
Keyword(3) | Error back-propagation |
Keyword(4) | Kullback's Measure |
1st Author's Name | Sumiyoshi Fujiki |
1st Author's Affiliation | GSIS, Tohoku University() |
2nd Author's Name | M. Nahomi Fujiki |
2nd Author's Affiliation | Sendai National College, of Technology |
3rd Author's Name | Mitsuyuki Nakao |
3rd Author's Affiliation | GSIS, Tohoku University |
Date | 1997/10/20 |
Paper # | NC97-43 |
Volume (vol) | vol.97 |
Number (no) | 332 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |