Summary of Global Graph Counterfactual Explanation: a Subgraph Mapping Approach, by Yinhan He et al.
Global Graph Counterfactual Explanation: A Subgraph Mapping Approach
by Yinhan He, Wendy Zheng, Yaochen Zhu, Jing Ma, Saumitra Mishra, Natraj Raman, Ninghao Liu, Jundong Li
First submitted to arxiv on: 25 Oct 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 This paper proposes a novel global-level graph counterfactual explanation method, called GlobalGCE, to address the limitations of existing local-level approaches. The goal is to identify a set of subgraph mapping rules that can change the Graph Neural Network (GNN) predictions for most graphs. To achieve this, the authors design a significant subgraph generator and a counterfactual subgraph autoencoder within their GlobalGCE framework. Experimental results demonstrate the superiority of their approach compared to existing baselines. The proposed method has implications for explaining GNNs in various real-world applications. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us understand how Graph Neural Networks (GNNs) can be better explained. Right now, most GNNs are like black boxes that don’t tell us why they made a certain prediction. One way to fix this is by finding small changes to the input graphs that would change the prediction. This is called counterfactual explanation. The problem with current methods is that they only work for individual graphs and don’t show us how different graphs are related. The new method, GlobalGCE, tries to solve this issue by finding rules that can be applied to many graphs at once. It’s like finding a formula that says “if you change these parts of the graph, it will make the prediction change in this way”. The authors tested their method and showed that it works better than other methods. |
Keywords
» Artificial intelligence » Autoencoder » Gnn » Graph neural network
