| Presentation | 2005-07-30 Neural networks learning for displaced feature Naomi INOUE, Kiyofumi KUSAKABE, Kouji HARA, Yoshimasa KIMURA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | This paper proposes the method which can represent the feature modification using neural networks. The neural networks can construct the model of the feature modification in itself if various patterns are presented as input patterns and reference pattern of the category is given as the desired signal and the learning is progressed. To clarify the internal representation formed in neural networks, two hidden units whose output variation is maximum and second maximum were investigated after the learning of Kanji character pattern "烏 (crow)" consisted of 100 samples were completed. As a result, the events were observed thet the hidden unit whose output variation is maximum has the relation with the horizontal degree and the bending degree of the third stroke and the sixth stroke in pattern "烏 (crow)", and the hidden unit whose output variation is second maximum has the relation with inclination information and the bending degree of the third stroke of "烏 (crow)". These results draw various feature modification are embeded in the hidden layer of neural networks. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Character recognition / feature displacement / neural networks / learning / internal representation |
| Paper # | MBE2005-50 |
| Date of Issue |
| Conference Information | |
| Committee | MBE |
|---|---|
| Conference Date | 2005/7/23(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 | ME and Bio Cybernetics (MBE) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Neural networks learning for displaced feature |
| Sub Title (in English) | |
| Keyword(1) | Character recognition |
| Keyword(2) | feature displacement |
| Keyword(3) | neural networks |
| Keyword(4) | learning |
| Keyword(5) | internal representation |
| 1st Author's Name | Naomi INOUE |
| 1st Author's Affiliation | Kochi University of Technology() |
| 2nd Author's Name | Kiyofumi KUSAKABE |
| 2nd Author's Affiliation | Kochi University of Technology |
| 3rd Author's Name | Kouji HARA |
| 3rd Author's Affiliation | Kochi University of Technology:(Present address)Currently, SUZUKI JIHAN KOCHI |
| 4th Author's Name | Yoshimasa KIMURA |
| 4th Author's Affiliation | Kochi University of Technology |
| Date | 2005-07-30 |
| Paper # | MBE2005-50 |
| Volume (vol) | vol.105 |
| Number (no) | 222 |
| Page | pp.pp.- |
| #Pages | 4 |
| Date of Issue | |