Summary of Invariant Graph Learning Meets Information Bottleneck For Out-of-distribution Generalization, by Wenyu Mao et al.
Invariant Graph Learning Meets Information Bottleneck for Out-of-Distribution Generalization
by Wenyu Mao, Jiancan Wu, Haoyang Liu, Yongduo Sui, Xiang Wang
First submitted to arxiv on: 3 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
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 framework, called Invariant Graph Learning based on Information bottleneck theory (InfoIGL), is proposed to enhance graph neural networks’ generalization ability under out-of-distribution (OOD) scenarios. InfoIGL extracts invariant features by compressing task-irrelevant information related to environmental factors using a redundancy filter. This approach maximizes mutual information among graphs of the same class in downstream classification tasks, preserving invariant features for prediction. The method achieves state-of-the-art performance on both synthetic and real-world datasets under OOD generalization for graph classification tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Graph learning can struggle with out-of-distribution (OOD) scenarios, where models perform poorly when dealing with new, unseen data. Researchers are exploring ways to improve this “generalization” ability. One approach is invariant learning, which focuses on extracting features that remain the same despite changes in the input data. This has worked well for images, but graph data is more complex and challenging. In this paper, a new method called InfoIGL is introduced to improve OOD generalization for graphs. It uses a combination of techniques, including a “redundancy filter” that removes unnecessary information, and multi-level contrastive learning to preserve important features. The result is a state-of-the-art approach that performs well on both synthetic and real-world datasets. |
Keywords
» Artificial intelligence » Classification » Generalization
