Summary of Achieving Dimension-free Communication in Federated Learning Via Zeroth-order Optimization, by Zhe Li et al.
Achieving Dimension-Free Communication in Federated Learning via Zeroth-Order Optimization
by Zhe Li, Bicheng Ying, Zidong Liu, Chaosheng Dong, Haibo Yang
First submitted to arxiv on: 24 May 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 The proposed Federated Learning (FL) framework, DeComFL, offers a novel dimension-free communication algorithm that reduces the communication cost from linear to constant, making it more efficient and suitable for large models. The algorithm leverages zeroth-order optimization techniques to transmit only scalar values between clients and the server in each round, regardless of the model’s dimension. This approach achieves state-of-the-art rates in non-convex functions, showing a linear speedup with the number of clients and local steps. Empirical evaluations demonstrate significant reductions in communication overhead, even when fine-tuning large language models. The code is available for further exploration. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary DeComFL is a new way to learn together on many devices without sharing all their data. This is helpful because sharing data can be risky. DeComFL helps by reducing the amount of information that needs to be shared. It does this by using special math tricks to only send small pieces of information between devices. This makes it much faster and more efficient. The results show that DeComFL can fine-tune large models with billions of parameters while only sending around 1MB of data. |
Keywords
» Artificial intelligence » Federated learning » Fine tuning » Optimization
