Summary of Fedgtst: Boosting Global Transferability Of Federated Models Via Statistics Tuning, by Evelyn Ma et al.
FedGTST: Boosting Global Transferability of Federated Models via Statistics Tuning
by Evelyn Ma, Chao Pan, Rasoul Etesami, Han Zhao, Olgica Milenkovic
First submitted to arxiv on: 16 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 Medium Difficulty Summary: The performance of Transfer Learning (TL) heavily relies on effective pretraining, which demands large datasets and substantial computational resources. Federated Learning (FL) addresses these issues by facilitating collaborations among clients, expanding the dataset indirectly, distributing computational costs, and preserving privacy. However, key challenges remain unresolved. Our proposed enhancements to FL aim to address these gaps. We introduce a client-server exchange protocol that leverages cross-client Jacobian norms to boost transferability and increase the average Jacobian norm across clients at the server as a local regularizer. This results in tighter control of the target loss, leading to an upper bound on the target loss in terms of the source loss and source-target domain discrepancy. Our transferable federated algorithm, FedGTST, demonstrates improved performance compared to relevant baselines, including FedSR and FedIIR, on datasets such as MNIST-M and SVHN when using LeNet as the backbone. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty Summary: This paper is about making it easier for people to use a type of artificial intelligence called Transfer Learning. Right now, it’s hard for one person or organization to do this by themselves because they need a lot of data and powerful computers. A way called Federated Learning helps with this by letting multiple people work together on the problem. However, there are still some challenges that haven’t been solved yet. The authors of this paper propose two new ideas to help fix these problems. They suggest exchanging information between different groups working together and using a special kind of “regularizer” to help make sure the results are good. Their new method, called FedGTST, works better than other methods on certain kinds of data. |
Keywords
* Artificial intelligence * Federated learning * Pretraining * Transfer learning * Transferability
