Summary of Fast-convergent and Communication-alleviated Heterogeneous Hierarchical Federated Learning in Autonomous Driving, by Wei-bin Kou et al.
Fast-Convergent and Communication-Alleviated Heterogeneous Hierarchical Federated Learning in Autonomous Driving
by Wei-Bin Kou, Qingfeng Lin, Ming Tang, Rongguang Ye, Shuai Wang, Guangxu Zhu, Yik-Chung Wu
First submitted to arxiv on: 29 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Robotics (cs.RO)
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 Street Scene Semantic Understanding (TriSU) framework addresses the challenge of autonomous driving in regions with diverse geographical settings. By leveraging Hierarchical Federated Learning (HFL), TriSU models can be trained collaboratively across distributed datasets from various cities. However, existing HFL methods struggle with slow convergence due to inter-city data heterogeneity. To tackle this issue, a Gaussian heterogeneous HFL algorithm (FedGau) is introduced, which models both RGB images and datasets as Gaussian distributions for aggregation weight design. This approach accelerates convergence by 35.5%-40.6% compared to state-of-the-art HFL methods. Additionally, a performance-aware adaptive resource scheduling (AdapRS) policy is proposed to minimize unnecessary communication overhead while maintaining performance. Extensive experiments demonstrate the effectiveness of AdapRS in reducing communication overhead by 29.65%. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper proposes a new way to make autonomous vehicles smarter. It solves a problem where AI models trained in one city don’t work well in another city because data from different cities is very different. The team developed a special algorithm called FedGau that helps the model learn faster and better by understanding the statistical patterns of each city’s data. This improvement allows the model to make decisions more quickly and accurately, even when it’s faced with new situations. To make this work efficiently, they also created an AdapRS policy that adjusts how much information is shared between cities based on how well the model is performing. This saves a lot of computational resources and makes the whole process faster. |
Keywords
» Artificial intelligence » Federated learning
