| Project/Area Number |
22KJ2289
|
|---|
| Project/Area Number (Other) |
22J11908 (2022)
|
|---|
| Research Category |
Grant-in-Aid for JSPS Fellows
|
| Allocation Type | Multi-year Fund (2023)
Single-year Grants (2022) |
|---|
| Section | 国内 |
|---|
| Review Section |
Basic Section 61030:Intelligent informatics-related
|
|---|
| Research Institution | Osaka University (2023)
Nara Institute of Science and Technology (2022) |
|---|
Principal Investigator |
Thonglek Kundjanasith 大阪大学, サイバーメディアセンター, 特任助教(常勤)
|
|---|
| Project Period (FY) |
2023-03-08 – 2024-03-31
|
| Project Status |
Completed (Fiscal Year 2023)
|
| Budget Amount *help |
¥1,700,000 (Direct Cost: ¥1,700,000)
Fiscal Year 2023: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 2022: ¥900,000 (Direct Cost: ¥900,000)
|
|---|
| Keywords | Distributed storage / Dependable computing / Edge machine learning / Privacy Preservation / Resource Heterogeneity / Swarm learning / Tactile Internet / Ultra-distributed system / Collaborative Develop / Distributed Computing / Edge Machine Learning / Federated Learning |
|---|
| Outline of Research at the Start |
I propose LiberatAI, an infrastructure for collaboratively developing machine learning models that allow researchers to work together. LiberatAI applies federated learning to train the models while preserving data privacy. LiberatAI allows individuals to collaboratively train models on their environments, which are usually heterogeneous. Three modules in LiberatAI support training a model on diverse storage, computing, and communication resources. LiberatAI was evaluated using the models to detect COVID-19 which is one of the most popular applications for privacy-sensitive data.
|
| Outline of Annual Research Achievements |
I have already finished developing my proposed infrastructure, LibearatAI. LibearatAI utilizes federated learning to train models while preserving data privacy. It enables individuals to collaboratively train models in their diverse environments, which typically vary. LibearatAI comprises three modules to support model training across different storage, computing, and communication resources: Furthermore, LibearatAI has been successfully applied to various applications, including hierarchical federated learning for predicting water levels in Thailand and decentralized federated learning for identifying plant diseases. These outcomes demonstrate LibearatAI's capacity to address challenges related to data sharing for data-driven services and extend its utility across diverse applications. In the future, LibearatAI is poised to become essential for ultra-distributed systems in the post-5G era. As the number of clients and data exponentially increases, its adaptability and privacy-preserving features will be crucial. LibearatAI's capacity to facilitate collaborative model training across diverse environments makes it a key solution for the challenges posed by escalating data demands.
|
