Analysis and learning of dynamic binary neural networks which can generate variable phenomena
Project/Area Number |
24500284
|
Research Category |
Grant-in-Aid for Scientific Research (C)
|
Allocation Type | Multi-year Fund |
Section | 一般 |
Research Field |
Sensitivity informatics/Soft computing
|
Research Institution | Hosei University |
Principal Investigator |
|
Project Period (FY) |
2012-04-01 – 2015-03-31
|
Project Status |
Completed (Fiscal Year 2014)
|
Budget Amount *help |
¥5,200,000 (Direct Cost: ¥4,000,000、Indirect Cost: ¥1,200,000)
Fiscal Year 2014: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2013: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2012: ¥2,080,000 (Direct Cost: ¥1,600,000、Indirect Cost: ¥480,000)
|
Keywords | ソフトコンピューティング / ニューラルネット / ニューラルネットワーク |
Outline of Final Research Achievements |
We have constructed a learning method to store one desired binary periodic orbit (BPO) into to the dynamic binary neural networks is presented. Applying the method to teacher signal BPOs that correspond to control signals of basic switching power converters, the efficiency of the method is confirmed. Introducing a digital return map, the dynamics of the DBNN is visualized and analyzed. In the case where a desired BPO can be stored into a DBNN, we have clarified that stability of the stored BPO can be reinforced (the number of initial points falling into the BPO is increased) by sparsifying connection matrix. In several basic examples of teacher signals, the efficiency of the sparsification is confirmed.
|
Report
(4 results)
Research Products
(32 results)