Summary of Axiomatic Explainer Globalness Via Optimal Transport, by Davin Hill et al.
Axiomatic Explainer Globalness via Optimal Transport
by Davin Hill, Josh Bone, Aria Masoomi, Max Torop, Jennifer Dy
First submitted to arxiv on: 2 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: 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 The paper proposes a new measure called Wasserstein Globalness to evaluate the diversity of explanations produced by feature attribution and feature selection methods. This complexity measure enables practitioners to compare and select explainers more effectively, as it provides insight into whether explanations are identical, unique, or somewhere in between. The authors define axiomatic properties that any such measure should possess and prove that Wasserstein Globalness meets these criteria. They validate the utility of this measure using various datasets, demonstrating its ability to facilitate meaningful comparisons between explainers and improve the selection process. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper explains a new way to compare and choose methods that help us understand how models make decisions. It’s hard to evaluate different methods because they produce different kinds of explanations. The authors introduce a new measure called Wasserstein Globalness, which helps us see if all the explanations are the same or unique in some way. They show that this measure is useful by testing it with different types of data and demonstrating how it can help us choose the best method for our needs. |
Keywords
» Artificial intelligence » Feature selection
