Fluctuation and Self-Organization in Nonlinear Systems
| Project/Area Number |
13640392
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
物性一般(含基礎論)
|
|---|
| Research Institution | Kokushikan University |
|---|
Principal Investigator |
SHIMIZU Toshihiro Kokushikan Univ., Faculty of Eng., Professor, 工学部, 教授 (30103641)
|
|---|
| Project Period (FY) |
2001 – 2002
|
| Project Status |
Completed (Fiscal Year 2002)
|
| Budget Amount *help |
¥1,600,000 (Direct Cost: ¥1,600,000)
Fiscal Year 2002: ¥1,000,000 (Direct Cost: ¥1,000,000)
Fiscal Year 2001: ¥600,000 (Direct Cost: ¥600,000)
|
|---|
| Keywords | Neural Network / Information Flow / Self-Organization / Pattern Formation / Synchronization between Neural Networks / Virtual Attractor / Associative Memory / Connection of Neural Networks / カオス / 階層構造 / 複雑系 / ブラウン運動 |
|---|
| Research Abstract |
To study the relation between fluctuation and self-organization in connected neural networks, we have first investigated a system of mutually coupled neural networks (Network A and Network B). Differeat patterns are stored in each network according to the Hebb rule and the associative memory problem is discussed in these networks. Since different patterns are stored in each network, the energy of each network has only the minimum corresponding to the patterns, if the networks were decoupled from each other. However, we have found that each network can retrieve the pattern stored in other network, if the networks are coupled. We think that a kind of attractor, corresponding to the pattern stored in another network, appears in each network. We call this attractor a virtual attractor. This result suggests that the information stored in one network can propagate to the other network
Next we have studied the self-organization processes in a global network, which consists of many connected local networks. As dynamics of neural network two cases are treated: (1) the chaotic Brownian network and (2) Chua-Yang model of CNN. In the first case it is found that various kinds of global patterns appear in the global system, if some patterns are selected autonomously by the system itself and they are stored in local networks. The global patterns are described in terms of 2 patterns stored in local networks as the background and the figure. In the second case it is investigated how the pattern (or information) stored hi a local network propagates in the global network and how the steady state pattern of the global network depends on the coupling between local networks. The relation between dynamics of the local network and the emergence of global patterns in the global network is discussed
|
Report
(3 results)
Research Products
(27 results)
