Summary of Derivative-free Diffusion Manifold-constrained Gradient For Unified Xai, by Won Jun Kim et al.
Derivative-Free Diffusion Manifold-Constrained Gradient for Unified XAI
by Won Jun Kim, Hyungjin Chung, Jaemin Kim, Sangmin Lee, Byeongsu Sim, Jong Chul Ye
First submitted to arxiv on: 22 Nov 2024
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 A novel explainability technique called Derivative-Free Diffusion Manifold-Constrainted Gradients (FreeMCG) is introduced to overcome shortcomings of gradient-based methods for image-based models. Traditional gradient-based methods require white-box access, are vulnerable to attacks, and produce attributions that lie off the image manifold. FreeMCG leverages ensemble Kalman filters and diffusion models to derive a derivative-free approximation of the model’s gradient projected onto the data manifold, requiring only output access. The method is demonstrated on counterfactual generation and feature attribution tasks, achieving state-of-the-art results while preserving essential XAI properties. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you want to understand how a computer program works or why it made a certain decision. This paper creates a new way to do that for images using neural networks. Right now, there are problems with existing methods because they require too much information and can be tricked by fake data. The new method, called FreeMCG, is better because it doesn’t need all the extra information and produces explanations that are more like what humans understand. |
Keywords
» Artificial intelligence » Diffusion
