Summary of Flea: Addressing Data Scarcity and Label Skew in Federated Learning Via Privacy-preserving Feature Augmentation, by Tong Xia and Abhirup Ghosh and Xinchi Qiu and Cecilia Mascolo
FLea: Addressing Data Scarcity and Label Skew in Federated Learning via Privacy-preserving Feature Augmentation
by Tong Xia, Abhirup Ghosh, Xinchi Qiu, Cecilia Mascolo
First submitted to arxiv on: 4 Dec 2023
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); 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 This paper proposes a novel Federated Learning (FL) framework called FLea, addressing challenges in existing FL methods when dealing with scarce and label-skewed data. FLea incorporates a global feature buffer to mitigate local model drift, a feature augmentation approach based on mix-ups for local training, and an obfuscation method to enhance privacy. Experimental results demonstrate FLea’s superiority over state-of-the-art FL counterparts in 13 out of 18 settings, with improvements of over 5%, while concurrently mitigating privacy vulnerabilities. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary FLea is a new way for devices to work together and learn without sharing their data. This helps solve problems when some devices have very little or imbalanced data, which makes it hard to train good models. FLea uses three main ideas: storing shared features from multiple devices, mixing up local training data to reduce overfitting, and hiding the source of shared features to keep them private. The results show that FLea works better than other FL methods in many cases. |
Keywords
* Artificial intelligence * Federated learning * Overfitting
