| Presentation | 1999/7/19 A Training Method For Mutilayer Nerual Networks Using Global Imformation of Error Surface Hiroyuki Nose, Kazuyuki Hara, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In this paper, we propose a training method that the convergence of the training does not depend on the initial connection weights. The proposed method uses global information of the error surface. BP algorithm is easily trapped to the local minima, because it uses only local information of the error surface. For this problem, we propose a training method which searches the network expresses the minimum output error by iterating a search process within a region on the error surface. Searching region becomes wider for a large output error and will be narrower for a small output error. This process allows the detail search of the error surface. Moreover, the proposed method updates only one connection weight at one searching process, so this works to reducing the computational complexity. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Global information / Error surface / Search / Annearing / Multilayer neural networks |
| Paper # | NC99-37 |
| Date of Issue |
| Conference Information | |
| Committee | NC |
|---|---|
| Conference Date | 1999/7/19(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 Training Method For Mutilayer Nerual Networks Using Global Imformation of Error Surface |
| Sub Title (in English) | |
| Keyword(1) | Global information |
| Keyword(2) | Error surface |
| Keyword(3) | Search |
| Keyword(4) | Annearing |
| Keyword(5) | Multilayer neural networks |
| 1st Author's Name | Hiroyuki Nose |
| 1st Author's Affiliation | Dept. Electrical Eng., Tokyo Metropolitan College of Technology() |
| 2nd Author's Name | Kazuyuki Hara |
| 2nd Author's Affiliation | Dept. Elec. &Info. Eng., Tokyo Metropolitan College of Technology |
| Date | 1999/7/19 |
| Paper # | NC99-37 |
| Volume (vol) | vol.99 |
| Number (no) | 193 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |