Summary of Feddistill: Global Model Distillation For Local Model De-biasing in Non-iid Federated Learning, by Changlin Song et al.
FedDistill: Global Model Distillation for Local Model De-Biasing in Non-IID Federated Learning
by Changlin Song, Divya Saxena, Jiannong Cao, Yuqing Zhao
First submitted to arxiv on: 14 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
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 introduces FedDistill, a novel framework for Federated Learning (FL) that addresses the issue of non-iid data distribution by enhancing knowledge transfer from the global model to local models. It tackles imbalanced class distribution through group distillation and separates the global model into feature extractors and classifiers to empower local models with more generalized data representation capabilities. The framework mitigates the adverse effects of data imbalance, ensuring that local models do not forget underrepresented classes but instead become more accurate at recognizing and classifying them. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated Learning is a new way for machines to learn together while keeping their data private. However, this approach has some problems when different devices have different types of data. The paper introduces a new idea called FedDistill that helps solve these issues by transferring knowledge from the global model to local models in a more balanced way. It also separates the global model into two parts: one that gets features and another that classifies them. This makes local models better at recognizing and classifying different types of data. |
Keywords
» Artificial intelligence » Distillation » Federated learning
