| 研究課題/領域番号 |
20K11764
|
|---|
| 研究機関 | 茨城大学 |
|---|
研究代表者 |
王 瀟岩 茨城大学, 理工学研究科(工学野), 准教授 (10725667)
|
|---|
| 研究分担者 |
梅比良 正弘 茨城大学, 理工学研究科(工学野), 特命研究員 (00436239)
|
|---|
| 研究期間 (年度) |
2020-04-01 – 2024-03-31
|
| キーワード | wireless access / federated learning |
|---|
| 研究実績の概要 |
In FY2022, we investigate the practical global model update process in wireless networks by proposing a robust asynchronized computing and communication process. Specifically, we proposed to decouple the computing and communication processes, and let the edge server use a subset of asynchronized local gradients to update the global model. We proved the algorithm’s convergence and evaluated its performance by simulations.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
2: おおむね順調に進展している
理由
We consider that the research progresses smoothly. We have published 2 journal papers, 2 internal conference papers and multiple domestic conference papers in FY2022 under the support of this funding.
|
| 今後の研究の推進方策 |
In the following year, we will implement the proposed layered in-network learning approach. Specifically, the main functions of FL will be implemented in the edge server (a GPU-workstation), and meanwhile, global gradient updating features would be added to the intelligent BSs. We will extensively evaluate the performance in terms of network throughput, spectrum utilization, bandwidth consumption and convergence speed, by using our developed testbed.
|
| 次年度使用額が生じた理由 |
Due to the affect of COVID19, most of the conferences are held online. Therefore, the grant originally planed for attending the conferences has not been used. These unused grant will be used as the participation fee for domestic and international conferences in FY2023.
|
