Theory and Implementation of Reservoir Computing Utilizing Non-Chaotic Strange Tractors
| Project/Area Number |
16KT0019
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 特設分野 |
|---|
| Research Field |
Mathematical Sciences in Search of New Cooperation
|
|---|
| Research Institution | Kyoto University |
|---|
Principal Investigator |
Aoyagi Toshio 京都大学, 情報学研究科, 教授 (90252486)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
中嶋 浩平 東京大学, 大学院情報理工学系研究科, 特任准教授 (10740251)
茶碗谷 毅 大阪大学, 情報科学研究科, 准教授 (80294148)
|
|---|
| Project Period (FY) |
2016-07-19 – 2020-03-31
|
| Project Status |
Completed (Fiscal Year 2019)
|
| Budget Amount *help |
¥18,200,000 (Direct Cost: ¥14,000,000、Indirect Cost: ¥4,200,000)
Fiscal Year 2018: ¥4,940,000 (Direct Cost: ¥3,800,000、Indirect Cost: ¥1,140,000)
Fiscal Year 2017: ¥6,110,000 (Direct Cost: ¥4,700,000、Indirect Cost: ¥1,410,000)
Fiscal Year 2016: ¥7,150,000 (Direct Cost: ¥5,500,000、Indirect Cost: ¥1,650,000)
|
|---|
| Keywords | レザバー計算 / カオス / 非線形力学系 / レザバー計算機 / ニューラルネットワーク / 情報量最大化 / ネットワーク結合力学系 / 力学系 / 数理工学 / 数理科学 / 神経科学 / 非線形 |
|---|
| Outline of Final Research Achievements |
In this study, to utilize large-degree-of-freedom nonlinear dynamical systems as a computational resource, we have explored appropriate guiding principles to construct a better Reservoir Computing and, at the same time, attempted to characterize the dynamical system itself as a new aspect of computational resources. First, we have examined the possibility of using dynamical systems that exhibit non-chaotic strange tractors as reservoirs, included the setting of appropriate tasks. Finally, focusing on the importance of the network structure that defines the interaction of the dynamical systems, we found that the guiding principle of recurrent mutual information maximization improves the performance, especially for the memory task.
|
| Academic Significance and Societal Importance of the Research Achievements |
レザバー計算は最近注目さているトピックであり、特に、大自由度非線形力学系の複雑な振る舞いを、計算資源として工学的応用に繋げる点は、数学の基礎理論と工学的応用の絡む学際的な研究テーマである．これまで、edge of chaos しか有効なレザバー構築の指針が無かったが、今回リカレント情報量最大化という新たな指針を示せたことは、工学的応用だけでなく、力学系の新たな特徴付けという基礎科学への貢献も期待出来る．最後に，数学的に非常にマニアックな力学系の性質が，情報処理システムとして工学的応用や脳・神経系の理解に関係している点は，基礎的学問の裾野の広い研究が如何に重要かを示す社会的意義も大きいといえる.
|
Report
(5 results)
Research Products
(75 results)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
[Presentation] Soft Keyboard: A Novel User Interface for Soft Devices2018
Author(s)
K. Nakajima, K. Inoue, Y. Kuniyoshi, S. Somlor, T. P. Tomo, A. Schmitz
Organizer
Special Session "Laser Dynamics and Complex Photonics 3: Reservoir Computing Session" in The 2018 International Symposium on Nonlinear Theory and Its Applications (NOLTA2018)
Related Report
Int'l Joint Research / Invited
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
