Summary of Balancing Label Imbalance in Federated Environments Using Only Mixup and Artificially-labeled Noise, by Kyle Sang et al.
Balancing Label Imbalance in Federated Environments Using Only Mixup and Artificially-Labeled Noise
by Kyle Sang, Tahseen Rabbani, Furong Huang
First submitted to arxiv on: 20 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 paper explores ways to address the issue of heterogeneous or non-iid federated learning in distributed environments, where data is skewed towards different subsets of labels. The authors propose a simple yet effective augmentation strategy that involves filling in underrepresented samples of a particular label class using pseudo-images. Unlike existing algorithms, this approach uses both real and pseudo-images in the client datasets. Additionally, the authors introduce two novel components: a DP-Instahide variant to reduce the decodability of image encodings and artificially labeled “natural noise” generated by an untrained StyleGAN. These noisy images mimic natural scene patterns, helping to homogenize label distribution among clients. The authors demonstrate that this approach can significantly improve training performance on CIFAR-10 and MNIST datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In a nutshell, this paper helps distributed learning systems work better when different users have very different data. It does this by creating fake images that look like the real ones but are more balanced in terms of what they’re supposed to be labeled as. This helps all the users learn from each other’s data more effectively. |
Keywords
» Artificial intelligence » Federated learning
