Summary of Cross-space Adaptive Filter: Integrating Graph Topology and Node Attributes For Alleviating the Over-smoothing Problem, by Chen Huang et al.
Cross-Space Adaptive Filter: Integrating Graph Topology and Node Attributes for Alleviating the Over-smoothing Problem
by Chen Huang, Haoyang Li, Yifan Zhang, Wenqiang Lei, Jiancheng Lv
First submitted to arxiv on: 26 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 A novel method is proposed to overcome the limitations of vanilla Graph Convolutional Networks (GCNs) in deep graph learning. GCNs use a low-pass filter to extract low-frequency signals from graph topology, which can lead to over-smoothing when going deep. To address this issue, various methods have been suggested that create an adaptive filter by incorporating an extra filter extracted from the graph topology. However, these methods rely heavily on topological information and ignore the node attribute space, sacrificing the expressive power of deep GCNs. The proposed Cross-Space Filter (CSF) addresses this limitation by producing adaptive-frequency information extracted from both topology and attribute spaces. CSF combines a tailored attribute-based high-pass filter with a topology-based low-pass filter, derived via multiple-kernel learning. This approach alleviates over-smoothing while promoting the effectiveness of node classification tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Deep learning on graphs can be limited by over-smoothing when using vanilla Graph Convolutional Networks (GCNs). A new method called Cross-Space Filter (CSF) helps solve this problem. GCNs use a low-pass filter to extract signals from graph topology, but this can cause over-smoothing as the network gets deeper. Some methods try to fix this by adding an extra filter that looks at the graph’s structure. However, these methods mostly focus on the graph’s structure and ignore information about each node. CSF combines two filters – one that looks at node attributes and another that looks at the graph’s structure. This helps solve over-smoothing while keeping deep GCNs effective. |
Keywords
* Artificial intelligence * Classification * Deep learning
