Summary of Accurate Explanation Model For Image Classifiers Using Class Association Embedding, by Ruitao Xie et al.
Accurate Explanation Model for Image Classifiers using Class Association Embedding
by Ruitao Xie, Jingbang Chen, Limai Jiang, Rui Xiao, Yi Pan, Yunpeng Cai
First submitted to arxiv on: 12 Jun 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- 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 generative explanation model combines global and local knowledge to explain image classifiers, addressing the limitations of existing methods that lack efficiency in extracting global knowledge. The class association embedding (CAE) method encodes each sample into a pair of separated class-associated and individual codes, enabling the generation of synthetic real-looking samples with modified class-associated features. A building-block coherency feature extraction algorithm efficiently separates class-associated features from individual ones, forming a low-dimensional manifold that visualizes classification decision patterns. The model can generate explanations for individual samples through counter-factual generation, continuously modifying the sample until its classification outcome changes. Compared to state-of-the-art methods, the proposed approach achieves higher accuracies in explaining image classification tasks using saliency maps. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study proposes a new way to explain how machines classify images. Existing models are like “black boxes” that don’t provide clear reasons for their decisions. The researchers develop a model that can generate explanations by changing individual features of an image until the class assignment changes. This approach is more accurate than existing methods and can help us understand why machines make certain classifications. |
Keywords
» Artificial intelligence » Classification » Embedding » Feature extraction » Image classification
