Summary of Goldfish: An Efficient Federated Unlearning Framework, by Houzhe Wang et al.
Goldfish: An Efficient Federated Unlearning Framework
by Houzhe Wang, Xiaojie Zhu, Chi Chen, Paulo Esteves-Veríssimo
First submitted to arxiv on: 4 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR)
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 paper proposes a novel framework, called Goldfish, for machine unlearning, which enables the removal of a user’s data from federated trained machine learning models without retraining from scratch. The framework consists of four modules: basic model, loss function, optimization, and extension. To address the challenge of low validity in existing algorithms, the paper introduces a novel loss function that considers the discrepancy between predictions and actual labels, as well as the bias of predicted results on the removed dataset. Additionally, the authors adopt knowledge distillation technique to enhance efficiency and introduce an early termination mechanism guided by empirical risk and data partition mechanism. The extension module incorporates adaptive distillation temperature and adaptive weight mechanisms to handle heterogeneity in user local data and quality of uploaded models. Experimental results demonstrate the effectiveness of the proposed approach. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper introduces a new way for machine learning models to “forget” specific data without needing to be retrained from scratch. This is important because people have the right to have their data removed from these models. The current methods for doing this are not very good, so the authors created a new framework called Goldfish. It has four parts: a basic model, a way to calculate how well it’s doing, an optimization part that makes it more efficient, and an extension that helps make sure the results are consistent. The new approach is better than the old one because it takes into account different things that can affect how well the model does. |
Keywords
» Artificial intelligence » Distillation » Knowledge distillation » Loss function » Machine learning » Optimization » Temperature
