H-FedSN: Personalized Sparse Networks for Efficient and Accurate Hierarchical Federated Learning for IoT Applications
Jiechao Gao, Yuangang Li *, Jie Wang, Michael Lepech, Yue Zhao, Brad Campbell · 2024
Read H-FedSN: Personalized Sparse Networks for Efficient and Accurate Hierarchical Federated Learning for IoT Applications by Jiechao Gao, Yuangang Li *, Jie Wang, Michael Lepech, Yue Zhao, Brad Campbell (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-06210.txt.
Description
Jiechao Gao and five co-authors published this paper in 2024. It tackles federated learning for the Internet of Things, where phones and sensors train a shared model without sending raw data away. Real device networks are layered and uneven, which strains the usual two-level designs.
Their system, H-FedSN, trims each device's network to a sparse personal version using an on-off mask, so far less data travels between layers. Shared layers keep a common base, while personal layers adapt to local data quirks. A Bayesian merging step at the edge and cloud levels weighs each group's input fairly. On three real device data sets, the system cut communication costs by up to 477 times against rivals. Accuracy stayed high, which suits low-power, slow-link settings.
Catalog record
| CREATOR | Jiechao Gao, Yuangang Li *, Jie Wang, Michael Lepech, Yue Zhao, Brad Campbell |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-06210