Adaptive collaboration for online personalized distributed learning with heterogeneous clients
Constantin Philippenko, Batiste Le Bars, Kevin Scaman, Laurent Massoulié · 2025
Read Adaptive collaboration for online personalized distributed learning with heterogeneous clients by Constantin Philippenko, Batiste Le Bars, Kevin Scaman, Laurent Massoulié (2025) 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-2507-06844.txt.
Description
Constantin Philippenko and co-authors tackle in a 2025 study how unlike clients in a shared learning network should pick partners. Each client trains its own model on its own data; borrowing gradients from similar peers cuts noise, but borrowing from unlike peers adds bias. Their rule watches gradient likeness and re-picks collaborators as training runs.
The rule grows from a sharper theory of an oracle scheme known to be optimal, and one variant keeps that optimality. Loss bounds cover smooth cases from strongly convex to rough non-convex shapes. Trials on synthetic and real data back the approach: adaptive partnering speeds local learning.
Catalog record
| CREATOR | Constantin Philippenko, Batiste Le Bars, Kevin Scaman, Laurent Massoulié |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-06844