Summary of Dual Feature-based and Example-based Explanation Methods, by Andrei V. Konstantinov et al.
Dual feature-based and example-based explanation methods
by Andrei V. Konstantinov, Boris V. Kozlov, Stanislav R. Kirpichenko, Lev V. Utkin
First submitted to arxiv on: 29 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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 proposed approach to local and global explanation selects a convex hull around an explained instance, generating a dual representation in the form of convex combinations of extreme points. Instead of perturbing new instances in Euclidean space, vectors of convex combination coefficients are generated from the unit simplex, forming a new dual dataset. A dual linear surrogate model is trained on this data, and feature importance values are computed using simple matrix calculations. This approach can be seen as a modification of LIME, and it inherently allows for example-based explanations. The neural additive model is also considered as a tool for implementing the approach. Numerical experiments with real datasets are performed to study the effectiveness. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new way to explain things is being explored. Instead of looking at individual instances in the same old way, this method combines them into a special shape called a convex hull. Then, it uses these combined instances to create a new dataset that’s easier to understand. A simple model is trained on this new data, and it helps us figure out which features are most important. This approach is similar to another popular explanation tool called LIME. It also allows us to get explanations based on specific examples. To test how well it works, many experiments were done using real datasets. |
