1992 Fiscal Year Final Research Report Summary
Modeling of function of biological internal fluctuation and its application to adaptive system
| Project/Area Number |
03680021
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Informatics
|
|---|
| Research Institution | Ibaraki University |
|---|
Principal Investigator |
AGU Masahiro INSTITUTION Ibaraki Univ. TITLE OF POSITION Prof. (Electronics), 工学部, 教授 (10016462)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
YAMANAKA Kazuo INSTITUTION Ibaraki Univ. TITLE OF POSITION Associated Prof. (Systems), 工学部, 助教授 (90125807)
|
|---|
| Project Period (FY) |
1991 – 1992
|
| Keywords | Fluctuation / Adaptability / Neural Network / Self-organization / Learning / 学習 / 自己組織 / ゾウリ虫 |
|---|
| Research Abstract |
The adaption scheme of " artificial biological particle" we proposed in this research is schematically shown in Fig.1. Here the particle is assumed to have a probabilistic internal source of fluctuation which controls the motion of the particle through actuator. The particle behaves like a Brownian particle moved by internal fluctuation. The sensored information of external field is used as an input to the source of spontaneous fluctuation in order to generate the control signal for the motion. Even if the input is fixed the output fluctuates because of the probabilistic property of the fluctuation source. The desired input-output probabilistic relation for given environmental field is gradually built by trial and error through the learning ability of the fluctuation source;
i) the adaptability of the motion indicated by a control is evaluated and feedback to the fluctuation source; ii) if a control gives rise to an adapted motion to external field, then the fluctuation source is recognized so as to generate the control more frequently. The more abundant the patterns of fluctuation are, the more flexible the adaptation of the particle are.
|
