Summary of Node-wise Filtering in Graph Neural Networks: a Mixture Of Experts Approach, by Haoyu Han et al.
Node-wise Filtering in Graph Neural Networks: A Mixture of Experts Approach
by Haoyu Han, Juanhui Li, Wei Huang, Xianfeng Tang, Hanqing Lu, Chen Luo, Hui Liu, Jiliang Tang
First submitted to arxiv on: 5 Jun 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 In this paper, researchers propose a new approach to improve Graph Neural Networks (GNNs) in handling complex graph structures by introducing a novel framework called Node-MoE. Traditional GNNs use a uniform global filter which can be suboptimal for real-world graphs that exhibit mixtures of homophilic and heterophilic patterns. The authors theoretically show that using a single global filter optimized for one pattern can harm performance on nodes with different patterns. To address this issue, Node-MoE utilizes a mixture of experts to adaptively select the appropriate filters for different nodes. The effectiveness of Node-MoE is demonstrated through extensive experiments on both homophilic and heterophilic graphs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps GNNs work better on complex graph structures. Researchers found that using one filter approach can hurt performance when dealing with mixed patterns in real-world graphs. They developed a new way called Node-MoE to choose the right filters for different nodes. This makes GNNs more effective for node classification tasks. |
Keywords
» Artificial intelligence » Classification » Mixture of experts
