Summary of Age-of-gradient Updates For Federated Learning Over Random Access Channels, by Yu Heng Wu et al.
Age-of-Gradient Updates for Federated Learning over Random Access Channels
by Yu Heng Wu, Houman Asgari, Stefano Rini, Andrea Munari
First submitted to arxiv on: 15 Oct 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 paper proposes a novel approach for federated training of deep neural networks over random access channels, which is crucial in computer networks, wireless networks, and cellular systems. The authors introduce a setting called RACH-FL, where remote users participate in training a centralized model using stochastic gradient descent under the coordination of a parameter server. They develop a policy called the “age-of-gradient” (AoG) policy that combines gradient sparsification, error correction, and slot transmission probability to address communication constraints between remote users and the parameter server. The AoG policy outperforms other RACH-FL policies in numerical simulations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper looks at how we can train deep learning models when many devices want to work together over a shared network. This is important for things like cellular networks or online communities where lots of people are sharing data and computing power. The authors come up with a new way to do this by combining different techniques to make sure that the devices don’t waste too much energy sending information back and forth. They test their method and show it works better than other approaches. |
Keywords
» Artificial intelligence » Deep learning » Probability » Stochastic gradient descent
