Summary of Verifying Relational Explanations: a Probabilistic Approach, by Abisha Thapa Magar et al.
Verifying Relational Explanations: A Probabilistic Approach
by Abisha Thapa Magar, Anup Shakya, Somdeb Sarkhel, Deepak Venugopal
First submitted to arxiv on: 5 Jan 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- 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 approach to verifying interpretable explanations generated by Graph Neural Networks (GNNs), specifically those produced by GNNExplainer. The existing method relies on human subjects, which is time-consuming and requires expertise. To scale up the verification process, the authors develop an uncertainty quantification framework that learns a factor graph model from counterfactual examples. These examples are generated as symmetric approximations of the relational structure in the original data. The approach is tested on several datasets, demonstrating its effectiveness in reliably estimating the uncertainty of relational explanations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us understand how to make sense of complex data relationships by checking if the reasons given for a prediction are correct. Usually, humans are used to verify these reasons because they don’t need special knowledge. However, verifying the quality of these reasons requires expertise and is hard to do on a large scale. The authors develop a new way to check the uncertainty of these reasons, using examples that reverse the relationships in the original data. This approach shows promising results on different datasets, allowing us to trust the explanations generated by GNNs. |
