PulseAugur
LIVE 16:59:47
research · [1 source] ·
0
research

Federated Global Reviser tackles noisy labels in machine learning

Researchers have introduced a new method called Federated Global Reviser (FedGR) to address the challenge of noisy labels in federated learning. This approach leverages the inherent property of global models in federated learning to slowly memorize noisy labels, enabling them to maintain reliable predictions. FedGR consists of three modules that work together to correct inaccurate labels and guide local training, significantly improving robustness against label noise and data heterogeneity. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Improves federated learning robustness against noisy labels and data heterogeneity, potentially enabling wider real-world application.

RANK_REASON This is a research paper introducing a novel method for federated learning with noisy labels.

Read on arXiv cs.CV →

COVERAGE [1]

  1. arXiv cs.CV TIER_1 · Yuxin Tian, Mouxing Yang, Yuhao Zhou, Jian Wang, Qing Ye, Tongliang Liu, Gang Niu, Jiancheng Lv ·

    Learning Locally, Revising Globally: Global Reviser for Federated Learning with Noisy Labels

    arXiv:2412.00452v2 Announce Type: replace-cross Abstract: Conventioanl federated learning (FL) heavily depends on high-quality labels, which are often impractical in the real world, leading to the federated label-noise (F-LN) problem. Worsely, the F-LN problem is exacerbated by t…