Summary of Rigorous Probabilistic Guarantees For Robust Counterfactual Explanations, by Luca Marzari et al.
Rigorous Probabilistic Guarantees for Robust Counterfactual Explanations
by Luca Marzari, Francesco Leofante, Ferdinando Cicalese, Alessandro Farinelli
First submitted to arxiv on: 10 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- 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 framework assesses the robustness of counterfactual explanations for deep learning models against plausible model shifts. By demonstrating the NP-completeness of computing robustness, the authors highlight the limitations of exact algorithms and instead propose a probabilistic approach that provides tight estimates with strong guarantees while preserving scalability. This method can analyze a wider range of architectures without imposing requirements on the network to be analyzed. The framework outperforms existing methods on four binary classification datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper studies how well deep learning models can explain their predictions when faced with changes in the model itself. They show that it’s hard to find a perfect solution for this problem, so they propose an alternative approach that’s fast and reliable. This new method can be used with many different types of neural networks without requiring special setup. The authors test their approach on four datasets and find that it does better than current methods in explaining model predictions. |
Keywords
* Artificial intelligence * Classification * Deep learning
