Summary of Explaining Black-box Model Predictions Via Two-level Nested Feature Attributions with Consistency Property, by Yuya Yoshikawa et al.
Explaining Black-box Model Predictions via Two-level Nested Feature Attributions with Consistency Property
by Yuya Yoshikawa, Masanari Kimura, Ryotaro Shimizu, Yuki Saito
First submitted to arxiv on: 23 May 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); 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 method is a model-agnostic local explanation technique that estimates high-level feature attributions (HiFAs) and low-level feature attributions (LoFAs) simultaneously for black-box machine learning models. This approach is crucial in increasing trust in AI systems by providing transparent explanations of the predictions. The nested structure of input features, comprising high- and low-level features, is exploited to estimate HiFAs and LoFAs that are consistent with each other. This consistency property bridges separate optimization problems for estimating HiFAs and LoFAs, allowing the proposed method to produce accurate and faithful explanations using a smaller number of queries to the models. The proposed method demonstrates its effectiveness in image classification in multiple instance learning and text classification using language models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper proposes a new way to explain how black-box AI models work. These models make predictions based on many different pieces of information, which can be hard to understand. The authors develop a technique that looks at both the big picture (high-level features) and the details (low-level features) to provide clear explanations. This helps increase trust in AI systems by making them more transparent. The method is tested on image and text classification tasks and shows promising results. |
Keywords
» Artificial intelligence » Image classification » Machine learning » Optimization » Text classification
