Summary of Dxai: Explaining Classification by Image Decomposition, By Elnatan Kadar et al.
DXAI: Explaining Classification by Image Decomposition
by Elnatan Kadar, Guy Gilboa
First submitted to arxiv on: 30 Dec 2023
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 DXAI method offers a novel approach to explaining and visualizing neural network classification through decomposition. Instead of providing an explanation heatmap, the technique yields a decomposition of the image into class-agnostic and class-distinct parts, relative to the data and chosen classifier. This leads to a fundamentally different way of understanding classification decisions. The decomposed parts can be combined to recreate the original image, offering insights into which features are responsible for classification. This new visualization can be particularly useful in scenarios where attributes are dense, global, and additive, such as when colors or textures are crucial for class distinction. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The DXAI method helps us understand how neural networks make decisions by breaking down an image into two parts: one that doesn’t give away the classification, and another that does. This can be super helpful in certain situations where it’s important to know which features of an image or data point are most relevant for a particular class. |
Keywords
* Artificial intelligence * Classification * Neural network
