Summary of Communication Efficient and Provable Federated Unlearning, by Youming Tao et al.
Communication Efficient and Provable Federated Unlearning
by Youming Tao, Cheng-Long Wang, Miao Pan, Dongxiao Yu, Xiuzhen Cheng, Di Wang
First submitted to arxiv on: 19 Jan 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 This paper tackles a novel problem in federated learning called federated unlearning. The authors aim to eliminate the impact of specific clients or data points on the global model learned via FL. This is driven by privacy concerns and the right to be forgotten. The authors introduce a new framework for exact federated unlearning, which meets two essential criteria: communication efficiency and exact unlearning provability. They develop a TV-stable FL algorithm called FATS, which modifies the classical FedAvg algorithm and employs local SGD with periodic averaging to lower communication rounds. The authors also design efficient unlearning algorithms under client-level and sample-level settings. They provide theoretical guarantees for their learning and unlearning algorithms, proving they achieve exact federated unlearning with reasonable convergence rates. Empirical results on 6 benchmark datasets show the framework’s superiority over state-of-the-art methods in terms of accuracy, communication cost, computation cost, and unlearning efficacy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making sure that when we learn from many different sources (like phones or computers), we can also forget some of that information if needed. This is important for privacy reasons and to respect people’s right to be forgotten. The authors created a new way to do this, called federated unlearning, which makes sure the learning process doesn’t hurt other parts of the model. They made an algorithm called FATS that can learn and forget in a way that is efficient and works well. |
Keywords
* Artificial intelligence * Federated learning
