Summary of Feddr+: Stabilizing Dot-regression with Global Feature Distillation For Federated Learning, by Seongyoon Kim et al.
FedDr+: Stabilizing Dot-regression with Global Feature Distillation for Federated Learning
by Seongyoon Kim, Minchan Jeong, Sungnyun Kim, Sungwoo Cho, Sumyeong Ahn, Se-Young Yun
First submitted to arxiv on: 4 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC); Machine Learning (cs.LG)
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 This paper addresses the challenge of client drift in Federated Learning (FL), where heterogeneous data distributions between clients hinder the aggregation of knowledge. Recent studies have tackled this issue by identifying divergence in the last classifier layer, but this approach may overemphasize observed classes within each client. The authors propose a novel algorithm, FedDr+, which enhances local alignment using dot-regression loss and improves aggregated global models through feature distillation to retain information about unseen/missing classes. By freezing the classifier as a simplex ETF, FedDr+ aligns features and boosts performance for both global and personalized FL. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about how we can make machine learning work better when different devices have different types of data. Right now, it’s hard to combine all that information because the data is so different. The authors are trying to fix this problem by creating a new way to learn from all those different devices at once. They’re doing this by freezing one part of the model and making sure everything else stays aligned. This will help us make better models that work well on lots of different devices. |
Keywords
» Artificial intelligence » Alignment » Distillation » Federated learning » Machine learning » Regression
