Dynamic Neural Architecture Warping for Time Series Recognition
Project/Area Number |
21K17808
|
Research Category |
Grant-in-Aid for Early-Career Scientists
|
Allocation Type | Multi-year Fund |
Review Section |
Basic Section 61030:Intelligent informatics-related
|
Research Institution | Kyushu University |
Principal Investigator |
|
Project Period (FY) |
2021-04-01 – 2023-03-31
|
Project Status |
Completed (Fiscal Year 2022)
|
Budget Amount *help |
¥3,900,000 (Direct Cost: ¥3,000,000、Indirect Cost: ¥900,000)
Fiscal Year 2022: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
Fiscal Year 2021: ¥2,730,000 (Direct Cost: ¥2,100,000、Indirect Cost: ¥630,000)
|
Keywords | Time Series / Neural Network / Dynamic Programming / Pattern Recognition / Machine Learning / Time series / Temporal neural network / Dynamic programming |
Outline of Research at the Start |
The purpose of this research is to address the existing issues with temporal neural networks, such as challenges with time series, inflexibility, lack of explainability, and problems with efficiency. To solve these problems, this research proposes the use of a novel neural architecture warping.
|
Outline of Final Research Achievements |
The research from this grant had wide successes in time series recognition, neural networks, and pattern recognition. Specifically, research wad done on dynamically warping the representations and the architectures of neural networks and applying it to new applications.
For research results, there were six peer-reviewed international journal publications, nine peer-reviewed international conference publications, and four Japanese conference publications. There are also three papers currently under review and two in the process of writing. All of these publications are high level journals and conferences.
|
Academic Significance and Societal Importance of the Research Achievements |
この研究は、ニューラルネットワークの頑健性と適用範囲の向上において重要です。多くの出版物は、時系列認識や文書認識の研究において長期的な影響を持ちます。
|
Report
(3 results)
Research Products
(22 results)