2020 Fiscal Year Final Research Report
Research on emergent design based on neuro-evolution with modularity
| Project/Area Number |
17H01795
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Soft computing
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
Iba Hitoshi 東京大学, 大学院情報理工学系研究科, 教授 (40302773)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
長谷川 禎彦 東京大学, 大学院情報理工学系研究科, 准教授 (20512354)
|
|---|
| Project Period (FY) |
2017-04-01 – 2020-03-31
|
| Keywords | 進化計算 / 遺伝的アルゴリズム / 遺伝的プログラミング / ニューロ進化 |
|---|
| Outline of Final Research Achievements |
In this study, we verified the usefulness of modular neuro-evolution. As a verification method, we analyzed whether a hierarchical repetitive structure can be obtained for the results of evolution applied to various benchmark problems. The modularity is expected to enable the reuse of information and the development of new structures during the temporal evolution of neural structures. In this way, we attempted to solve the difficulties associated with the design of conventional artifacts (i.e., the inability to obtain an appropriate overall solution from partial solutions and the inability to adapt robustly to changes in the environment).
|
| Free Research Field |
進化計算と知能創発
|
| Academic Significance and Societal Importance of the Research Achievements |
これまで神経回路網の進化と時間発展の考えを取り入れて構築されたモデルはほとんどない．さらにこの統一的な視点にたって人工物設計での実証研究を試みるものは極めて少ない．そのため，本研究は理論的基盤にたってニューロ進化の枠組みをロバストな設計に応用するという学術的な特色がある．本研究により，ニューロ進化による知能創発の有効性が，ハードウェアとソフトウェアの創発デザインを通して明らかになる．
|
