Summary of Structure Your Data: Towards Semantic Graph Counterfactuals, by Angeliki Dimitriou et al.
Structure Your Data: Towards Semantic Graph Counterfactuals
by Angeliki Dimitriou, Maria Lymperaiou, Giorgos Filandrianos, Konstantinos Thomas, Giorgos Stamou
First submitted to arxiv on: 11 Mar 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- 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 The proposed counterfactual explanation (CE) method leverages semantic graphs accompanying input data to generate more descriptive, accurate, and human-aligned explanations. Building upon state-of-the-art conceptual attempts, the model-agnostic edit-based approach introduces Graph Neural Networks (GNNs) for efficient Graph Edit Distance (GED) computation. The method represents images as scene graphs and obtains their GNN embeddings to bypass solving the NP-hard graph similarity problem for all input pairs. Experiments on benchmark and real-world datasets with varying difficulty and availability of semantic annotations show that the proposed CEs outperform previous state-of-the-art explanation models based on semantics, including both white-box and black-box as well as conceptual and pixel-level approaches. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Our new way of explaining why computers make certain decisions is called counterfactual explanations (CEs). We’re using special graphs to help us understand which important features in the data led to a particular decision. This is different from other methods that just look at individual pixels or concepts. Our approach is better because it takes into account the relationships between these features. We tested our method on different types of images and classifiers, and people found our explanations more helpful and accurate. |
Keywords
» Artificial intelligence » Gnn » Semantics
