2022 Fiscal Year Final Research Report
Scaling up CNN computations for data-intensive scientific applications
| Project/Area Number |
20K19823
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 61010:Perceptual information processing-related
|
|---|
| Research Institution | Kobe University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Keywords | Deep Learning 4 science / Computational Efficiency / Computer Vision / ConvNets |
|---|
| Outline of Final Research Achievements |
In this research, our objective has been to develop new computational tools that can be applied to tackle diverse scientific and engineering problems. We have focused our efforts on devising methods to improve the resource consumption of deep learning models.
To demonstrate the effectiveness of our algorithms, we have benchmarked them on a wide array of scientific applications across the fields of neuro-science, biodiversity monitoring and material science.
Beyond the tangible improvements in the computational efficiency, our work has also opened up new avenues for interdisciplinary collaboration and innovation: Our software has been used to help an Israeli startup prototype low-level vision models and, in partnership with the Paris Observatory, we have been developing efficient hydrological models to improve our understanding of water resource management and sustainability.
|
| Free Research Field |
機械学習・深層学習
|
| Academic Significance and Societal Importance of the Research Achievements |
深層学習は、現行の技術では手が届かないとされていた技術的な課題への解決策を開いてきましたが大量の計算力と高額なインフラが必要となるため、技術の開発と応用は大規模な技術機関内に大きく集中しています。したがって、計算効率を改善し、その応用の利益を広範囲の人々に広げることが必要となっています。本研究では、限定的な計算力で展開できる技術を開発し、その適用性を複数の実用的な科学的な応用例を通じて示しました。
|
