Summary of Decoupled Federated Learning on Long-tailed and Non-iid Data with Feature Statistics, by Zhuoxin Chen et al.
Decoupled Federated Learning on Long-Tailed and Non-IID data with Feature Statistics
by Zhuoxin Chen, Zhenyu Wu, Yang Ji
First submitted to arxiv on: 13 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 Decoupled Federated learning framework using Feature Statistics (DFL-FS) addresses the challenges of training models on heterogeneous data in long-tailed and non-IID distributions. By estimating client class coverage distributions through masked local feature statistics clustering, DFL-FS selects models for aggregation to accelerate convergence and enhance feature learning without compromising privacy. The framework employs federated feature regeneration based on global feature statistics and resampling/weighted covariance to calibrate the global classifier, enhancing adaptability to long-tailed data distributions. Experimental results on CIFAR10-LT and CIFAR100-LT datasets show that DFL-FS outperforms state-of-the-art methods in terms of accuracy and convergence rate. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated learning helps keep data private, but it has a problem when dealing with different kinds of data. This paper looks at what happens when some types of data are very rare and scattered across just a few places. As a result, the models trained on these data don’t get used much during training, which slows down how well the model learns. To fix this, the authors suggest using two steps: first, they estimate how many times each type of data appears at each place, then use that information to choose which models to combine and speed up learning. This helps both by making sure all types of data are included and by keeping the model’s performance good. |
Keywords
* Artificial intelligence * Clustering * Federated learning
