| Presentation | 2002/3/11 Ability of Prediction and Learning of Digital Neural Network with Specific Connections Masahiro ISHIDO, Hideki YOSHIDA, Kousuke TAMADA, Shuhei TAKAKI, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In the hierarchical digital neural network which consists of FIR filter being connected with delay elements in accordance with a geometric configuration, the learning of prediction (1, 3 or 5 steps ahead) was done by using GA and a sinusoidal wave with frequency fo as the learning wave. After the learning, the network has good ability of prediction for any waves with frequency less than fo. The prediction ratio at one step ahead was 99.1 %. To examine the reason why the network has the ability of prediction for any waves by the learning of only one sinusoidal wave, the ability of prediction for sinusoidal waves with frequency from 0 through fo was examined. According to the result, the prediction ratio at five steps ahead is more than 90.0 % all over the frequency range. It is considered that the network recognizes the frequencies components, gives the prediction of each sinusoidal wave of the Fourier components and adds the predicted values to get the prediction of the signals. The possibility of prediction of differentiated value of time serial signals by using the digital neural network was also examined. It resulted that the network showed good ability for differential prediction of any waves with frequency components less than fo. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Digital Neural Network / FIR Filter / Prediction / Differential Prediction |
| Paper # | NC2001-134 |
| Date of Issue |
| Conference Information | |
| Committee | NC |
|---|---|
| Conference Date | 2002/3/11(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) | Ability of Prediction and Learning of Digital Neural Network with Specific Connections |
| Sub Title (in English) | |
| Keyword(1) | Digital Neural Network |
| Keyword(2) | FIR Filter |
| Keyword(3) | Prediction |
| Keyword(4) | Differential Prediction |
| 1st Author's Name | Masahiro ISHIDO |
| 1st Author's Affiliation | Department bof Computer and sysytems Engineering, Faculty of Engineering, Kobe University() |
| 2nd Author's Name | Hideki YOSHIDA |
| 2nd Author's Affiliation | Department bof Computer and sysytems Engineering, Faculty of Engineering, Kobe University |
| 3rd Author's Name | Kousuke TAMADA |
| 3rd Author's Affiliation | Department bof Computer and sysytems Engineering, Faculty of Engineering, Kobe University |
| 4th Author's Name | Shuhei TAKAKI |
| 4th Author's Affiliation | Department bof Computer and sysytems Engineering, Faculty of Engineering, Kobe University |
| Date | 2002/3/11 |
| Paper # | NC2001-134 |
| Volume (vol) | vol.101 |
| Number (no) | 735 |
| Page | pp.pp.- |
| #Pages | 8 |
| Date of Issue | |