Summary of Hgnas: Hardware-aware Graph Neural Architecture Search For Edge Devices, by Ao Zhou et al.
HGNAS: Hardware-Aware Graph Neural Architecture Search for Edge Devices
by Ao Zhou, Jianlei Yang, Yingjie Qi, Tong Qiao, Yumeng Shi, Cenlin Duan, Weisheng Zhao, Chunming Hu
First submitted to arxiv on: 23 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 Graph Neural Networks (GNNs) excel in graph-based learning tasks, particularly point cloud processing. However, the research community has neglected efficient GNN design for edge scenarios with real-time requirements and limited resources. This paper proposes HGNAS, a novel hardware-aware graph neural architecture search framework tailored for resource-constrained edge devices. To achieve hardware awareness, HGNAS integrates an efficient GNN hardware performance predictor that evaluates latency and peak memory usage in milliseconds. Additionally, it offers a peak memory estimation method to enhance robustness during inference. The framework constructs a fine-grained design space to explore extreme-performance architectures by decoupling the GNN paradigm. A multi-stage hierarchical search strategy navigates huge candidates, reducing single-search time to a few GPU hours. To the best of our knowledge, HGNAS is the first automated GNN design framework for edge devices and achieves hardware awareness across different platforms. Extensive experiments demonstrate HGNAS’ superiority, achieving up to 10.6x speedup and 82.5% peak memory reduction with negligible accuracy loss compared to DGCNN on ModelNet40. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making computer models that work well on devices like smartphones or smart home gadgets. These devices have limited resources, so we need a way to make the models run faster and use less memory. The researchers created a new system called HGNAS that can design these models efficiently for edge devices. It uses special tools to predict how much time and memory each model will need and makes sure it’s running smoothly. This is important because current systems often run out of memory or take too long to process information, which slows them down. The researchers tested their system on different devices and found that it works well, making the models up to 10 times faster and using less than 82% more memory. |
Keywords
» Artificial intelligence » Gnn » Inference
