Context-Aware Online Model Splitting and Device Association for Semi-Decentralized Federated Learning in Internet of Things
Bo Xu, Shuang Wang, Xiaoyu Tang · 2026
Read Context-Aware Online Model Splitting and Device Association for Semi-Decentralized Federated Learning in Internet of Things by Bo Xu, Shuang Wang, Xiaoyu Tang (2026) 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/PMC13364252.txt.
Description
Bo Xu, Shuang Wang, and Xiaoyu Tang present a 2026 study on training AI across Internet of Things devices by blending federated and split learning without a single central server. In their semi-decentralized design, devices share the load of big models, split at points that fit each device's power and link, and several edge servers aggregate the shared parts. Since training delay is unknown ahead of time, they treat split and pairing choices as a learning problem in itself.
Their context-aware bandit algorithm reads each site's conditions to guess delays, then improves its choices round by round with provably small long-run regret. Experiments on several networks showed lower training latency and better test accuracy than the rival methods chosen. The work targets privacy-keeping training on weak devices.
Catalog record
| CREATOR | Bo Xu, Shuang Wang, Xiaoyu Tang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64884 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |