| Presentation | 2002/12/6 A Neural Network With Variable Connection Weights and Its Application to Function Approximation Makoto KOURIN, Kenji NAKAYAMA, Akihiro HIRANO, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Boundary surface of a neuron with a linear combination of inputs is a hyper-plane. Multilayer neural networks (NN) usually compose many the hyper-planes, and form a complicated boundary. In this paper, a variable connection weight is proposed. In the linear combination of the inputs, the connection weight from the bias to the neuron is controlled by the input. A single neuron with this variable connection weight can form a curved surface boundary. A multi-stage learning algorithm is proposed for the NN with the variable connection weights. In the first step, the NN with the normal connection weights, which are not controlled by the inputs, is trained. Next,the variable connection weights are adjusted using the back-propagated error. Finally, both connection weights are simultaneously trained. The learning is based on the error back-propagation algorithm. Simulation results show the proposed NN can approximate complicated functions using a small number of neurons. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Neural networks / Boundary surface / Variable connection weights / Learning / Function approximation |
| Paper # | NC2002-96 |
| Date of Issue |
| Conference Information | |
| Committee | NC |
|---|---|
| Conference Date | 2002/12/6(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 Neural Network With Variable Connection Weights and Its Application to Function Approximation |
| Sub Title (in English) | |
| Keyword(1) | Neural networks |
| Keyword(2) | Boundary surface |
| Keyword(3) | Variable connection weights |
| Keyword(4) | Learning |
| Keyword(5) | Function approximation |
| 1st Author's Name | Makoto KOURIN |
| 1st Author's Affiliation | Division of Electironics and Computer Science, Graduate School of Natural Science and Technology Kanazawa University() |
| 2nd Author's Name | Kenji NAKAYAMA |
| 2nd Author's Affiliation | Dept. of Information and Systems Eng. Faculty of Eng. Kanazawa University |
| 3rd Author's Name | Akihiro HIRANO |
| 3rd Author's Affiliation | Dept. of Information and Systems Eng. Faculty of Eng. Kanazawa University |
| Date | 2002/12/6 |
| Paper # | NC2002-96 |
| Volume (vol) | vol.102 |
| Number (no) | 508 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |