Summary of Fedrepopt: Gradient Re-parametrized Optimizers in Federated Learning, by Kin Wai Lau et al.
FedRepOpt: Gradient Re-parametrized Optimizers in Federated Learning
by Kin Wai Lau, Yasar Abbas Ur Rehman, Pedro Porto Buarque de Gusmão, Lai-Man Po, Lan Ma, Yuyang Xie
First submitted to arxiv on: 24 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computer Vision and Pattern Recognition (cs.CV); 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 Federated Learning (FL) enables training machine learning models on edge devices while preserving privacy. However, as model sizes increase, gradient updates become less frequent due to computational power and memory limitations. This results in suboptimal training outcomes during FL rounds, limiting the feasibility of deploying advanced models on edge devices. To address this issue, we propose FedRepOpt, a gradient re-parameterized optimizer for FL. By modifying gradients according to model-specific hyperparameters obtained from complex models, FedRepOpt enables simple local models to achieve similar performance as complex ones. In this work, we focus on VGG-style and Ghost-style models in the FL environment. Our experiments demonstrate that models using FedRepOpt obtain a significant boost in performance (16.7% and 11.4%) while also showing faster convergence times (11.7% and 57.4%) compared to their complex counterparts. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated Learning is a way to train machine learning models on devices without sharing private data. But, as models get bigger, they use less power and memory, making them slower to learn. This makes it harder to use big models on small devices. To solve this problem, we created FedRepOpt, a new way to optimize model updates in FL. Our method helps simple local models perform just as well as complex ones by adjusting how gradients are calculated. In our tests, we used two types of models (VGG-style and Ghost-style) and found that using FedRepOpt gave us better results and faster training times. |
Keywords
» Artificial intelligence » Federated learning » Machine learning
