Summary of Addressing Skewed Heterogeneity Via Federated Prototype Rectification with Personalization, by Shunxin Guo et al.
Addressing Skewed Heterogeneity via Federated Prototype Rectification with Personalization
by Shunxin Guo, Hongsong Wang, Shuxia Lin, Zhiqiang Kou, Xin Geng
First submitted to arxiv on: 15 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Distributed, Parallel, and Cluster Computing (cs.DC)
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 proposed Federated Prototype Rectification with Personalization (FPRP) framework tackles data-level heterogeneity in federated learning by introducing two novel components: Federated Personalization and Federated Prototype Rectification. These components aim to balance decision boundaries between dominant and minority classes using private data, while also rectifying empirical prototypes through inter-class discrimination and intra-class consistency. The approach is evaluated on three popular benchmarks, outperforming current state-of-the-art methods and achieving balanced performance in both personalization and generalization. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated learning helps devices train models together without sharing their data. A problem with this is that some devices might have very different data. This can make it hard for the model to work well on all devices. Researchers are working on ways to solve this problem, but most methods assume that the data is spread evenly across all devices. In reality, the data might be skewed or unevenly distributed. This paper looks at this challenge and proposes a new way to address it called Skewed Heterogeneous Federated Learning (SHFL). The approach involves two parts: one helps construct balanced decision boundaries, and the other rectifies prototypes by looking at both class differences and similarities. The method is tested on three datasets and outperforms current methods. |
Keywords
» Artificial intelligence » Federated learning » Generalization
