Dynamics of Higher Order Neural Networks and its Application
| Project/Area Number |
10650372
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
情報通信工学
|
|---|
| Research Institution | Kagoshima University |
|---|
Principal Investigator |
MIYAJIMA Hiromi Kagoshima University, Fac. of Eng., Professor, 工学部, 教授 (60132669)
|
|---|
| Project Period (FY) |
1998 – 1999
|
| Project Status |
Completed (Fiscal Year 1999)
|
| Budget Amount *help |
¥2,000,000 (Direct Cost: ¥2,000,000)
Fiscal Year 1999: ¥900,000 (Direct Cost: ¥900,000)
Fiscal Year 1998: ¥1,100,000 (Direct Cost: ¥1,100,000)
|
|---|
| Keywords | Neural networks / Higher Order Model / Multilayer Perceptron / Associative Memery / Dynamics / Prediction of Time Series / Diagnosis System of Injury by Salt / Pattern Recognition / 高次ニューラルネットワーク / パターン分離 / ランダム神経回路網 |
|---|
| Research Abstract |
In this research with higher order neural networks(HONN), we have obtained the following results.
(1) HONN are superior in various numerical simulations to the conventional model.
(1-1) HONN are superior in function approximation, pattern recognition and prediction of time series by using the same number of parameters of each model to the conventional model.
(1-2) In the case of determining the structure of networks, HONN are superior to the conventional model.
(1-3) HONN are superior in associative memory of sequential patterns to the conventional model.
(2) Theoretical analysis are made with dynamic of HONN.
(2-1) Dynamics of recalling ability in associative memory is shown theoretically.
(2-2) It is shown that HONN are superior in separation ability of patterns to the conventional model.
(3) HONN are superior in separation to the conventional model.
Diagnosis systems of injury by salt for distribution lines are constructed by neural networks.
It is shown that HONN are superior in this problem to the conventional model.
|
Report
(3 results)
Research Products
(26 results)
