Percolative Learning and its applications
Project/Area Number |
18H03305
|
Research Category |
Grant-in-Aid for Scientific Research (B)
|
Allocation Type | Single-year Grants |
Section | 一般 |
Review Section |
Basic Section 61040:Soft computing-related
|
Research Institution | Yokohama National University |
Principal Investigator |
Nagao Tomoharu 横浜国立大学, 大学院環境情報研究院, 教授 (10180457)
|
Project Period (FY) |
2018-04-01 – 2021-03-31
|
Project Status |
Completed (Fiscal Year 2020)
|
Budget Amount *help |
¥17,550,000 (Direct Cost: ¥13,500,000、Indirect Cost: ¥4,050,000)
Fiscal Year 2020: ¥4,680,000 (Direct Cost: ¥3,600,000、Indirect Cost: ¥1,080,000)
Fiscal Year 2019: ¥4,680,000 (Direct Cost: ¥3,600,000、Indirect Cost: ¥1,080,000)
Fiscal Year 2018: ¥8,190,000 (Direct Cost: ¥6,300,000、Indirect Cost: ¥1,890,000)
|
Keywords | 機械学習 / 深層学習 / ニューラルネットワーク / 進化計算法 / マルチモーダル / 時系列予測 / マルチモーダル学習 |
Outline of Final Research Achievements |
We previously developed "PLM: Percolative Learning Method" which is a kind of learning method for layered deep neural networks. In this project, we studied theory, methods and applications of PLM. In PLM, we can "percolate" Aux data which are used only for learning into Main data which are used for learning and testing. We proved that PLM could achieve the precision rate higher than a conventional deep neural network experimentally. We dealt with time dependence data prediction and realization of software-sensor, and we showed that PLM is effective for various fields.
|
Academic Significance and Societal Importance of the Research Achievements |
これまでの神経回路網や深層学習では,学習時のみ利用できるデータは,学習しても実際の運用の際に使えなくなるので,結局利用されてこなかった.これに対して,我々が開発した浸透学習を使うことで,そのようなデータを有効に活用することができるようになった.本事業において浸透学習の理論・方法・応用について具体的な検討を行い,浸透学習が有効であることを示すことができた.今後,様々な分野で浸透学習を利用することが考えられ,その効果と波及効果は大きいと考えられる.
|
Report
(4 results)
Research Products
(3 results)