Summary of Respect the Model: Fine-grained and Robust Explanation with Sharing Ratio Decomposition, by Sangyu Han et al.
Respect the model: Fine-grained and Robust Explanation with Sharing Ratio Decomposition
by Sangyu Han, Yearim Kim, Nojun Kwak
First submitted to arxiv on: 25 Jan 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 eXplainable AI (XAI) method, called SRD (Sharing Ratio Decomposition), addresses concerns over the truthfulness of existing explanation methods. By sincerely reflecting the model’s inference process, SRD achieves significantly enhanced robustness in explanations. Unlike conventional approaches focusing on neuronal levels, SRD considers intricate nonlinear interactions between filters using a vector perspective. The method also introduces an interesting observation called Activation-Pattern-Only Prediction (APOP), redefining relevance to include both active and inactive neurons. This allows for the recursive decomposition of Pointwise Feature Vectors (PFVs) and provides high-resolution Effective Receptive Fields (ERFs) at any layer. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper proposes a new way to understand how artificial intelligence models make decisions. The current methods are not very accurate, so they can be tricked into giving wrong answers. To fix this, the authors developed a new method called SRD that shows exactly how the model is thinking. Instead of looking at individual parts, SRD looks at the connections between them. This helps to identify important features and show why certain decisions were made. |
Keywords
» Artificial intelligence » Inference
