2013 Fiscal Year Final Research Report
Robust training based on combined online/batch training techniques
| Project/Area Number |
23500189
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Shonan Institute of Technology |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Renkei-kenkyūsha) |
KOBAYASHI Manabu 湘南工科大学, 工学部, 教授 (80308204)
|
|---|
| Project Period (FY) |
2011 – 2013
|
| Keywords | ニューラルネットワーク / 学習アルゴリズム / 準ニュートン法 / オンライン学習 / バッチ学習 / 並列アルゴリズム |
|---|
| Research Abstract |
In this research, it is a purpose to enable the approximation model by the feedforward neural networks for the function or the system with the highly nonlinear behavior by the following studies. Specifically, "Proposal of a novel training algorithm using combined online/batch quasi-Newton techniques", and "Analysis on the robustness of the proposed algorithm". Here, robustness in this research means that the proposed algorithm has strong ability to search a global minimum without being trapped into local minimum. Furthermore, this approach is useful for the circuit modeling for the design and optimization, where analytical formulas are not available or original model is computationally too expensive. A neural model is trained once, and can be used again and again. This avoids repetitive circuit simulations where a change in the physical dimension requires a re-simulation of the circuit structure.
|
