Summary of Learning Locally, Revising Globally: Global Reviser For Federated Learning with Noisy Labels, by Yuxin Tian et al.
Learning Locally, Revising Globally: Global Reviser for Federated Learning with Noisy Labels
by Yuxin Tian, Mouxing Yang, Yuhao Zhou, Jian Wang, Qing Ye, Tongliang Liu, Gang Niu, Jiancheng Lv
First submitted to arxiv on: 30 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The study proposes a novel approach called Global Reviser for Federated Learning with Noisy Labels (FedGR) to enhance the label-noise robustness of federated learning. The authors observe that global models in FL slowly memorize noisy labels, which is problematic in real-world scenarios like medicine where accurate labels are often inaccessible. FedGR employs three novel modules: noisy label sniffing and refining, local knowledge revising, and local model regularization. These modules utilize the global model to infer local data proxies, refine incorrect labels, revise local models, and regularize overfitting. The authors demonstrate the effectiveness of FedGR through extensive experiments on three F-LNL benchmarks against seven baseline methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In this study, researchers try to solve a big problem with federated learning called the “federated label-noise” (F-LN) problem. Federated learning is when many devices or machines learn together without sharing their data. But often, these devices don’t have accurate labels for what they’re trying to learn. The authors of this study propose a new way to make federated learning work better in these situations. They call it “Global Reviser for Federated Learning with Noisy Labels” (FedGR). It uses special modules that help the global model figure out which labels are correct and which aren’t. |
Keywords
» Artificial intelligence » Federated learning » Overfitting » Regularization
