FedPAW: Federated Learning with Personalized Aggregation Weights for Urban Vehicle Speed Prediction
Yuepeng He, Pengzhan Zhou, Yijun Zhai, Fang Qu, Zhida Qin, Mingyan Li, Songtao Guo · 2024
Read FedPAW: Federated Learning with Personalized Aggregation Weights for Urban Vehicle Speed Prediction by Yuepeng He, Pengzhan Zhou, Yijun Zhai, Fang Qu, Zhida Qin, Mingyan Li, Songtao Guo (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-2412-01281.txt.
Description
Yuepeng He and six co-authors predict city traffic speeds vehicle by vehicle, without any driver's data leaving the car, in this 2024 paper. Federated learning trains a shared model across many clients who keep their data local. But drivers differ, in style and in vehicle, so one global model fits nobody particularly well.
Their framework, FedPAW, personalizes the federation. Each client's model is compared with the global one, and the mismatch, measured as weighted error, becomes a set of personal aggregation weights. The server then sends each client a tailored blend instead of the identical model, at no extra computing or communication cost to the client. In tests on driving data gathered in the CARLA simulator, with a sequence model reading speed histories, the personalized federation predicted future speeds best. Privacy and personalization, often traded against each other, here arrive together.
Catalog record
| CREATOR | Yuepeng He, Pengzhan Zhou, Yijun Zhai, Fang Qu, Zhida Qin, Mingyan Li, Songtao Guo |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-01281