Federated Learning and Class Imbalances: Robust Heterogeneous Federated Learning with Label Noise
Siqi Zhu, Joshua D. Kaggie · 2026
Read Federated Learning and Class Imbalances: Robust Heterogeneous Federated Learning with Label Noise by Siqi Zhu, Joshua D. Kaggie (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/arxiv-2601-06348.txt.
Description
Training models across hospitals without moving patient data is the promise of federated learning. Siqi Zhu and 1 co-author stress-test a leading method under class imbalance in a 2026 paper. Real deployments face label noise and data that differ sharply between sites, which can quietly wreck performance.
Their work reproduces RHFL+, a method designed for clients with different model architectures, alongside all its benchmark rivals under one evaluation framework. They extend it to real medical imaging datasets covering breast imaging and histopathology. A new implementation uses NVIDIA's production federated framework, making the codebase modular and deployment-ready. Ablation studies, noise comparisons, and scaling tests with growing client counts check whether the method's robustness holds up in practice.
Catalog record
| CREATOR | Siqi Zhu, Joshua D. Kaggie |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-06348