Summary of Interpret the Predictions Of Deep Networks Via Re-label Distillation, by Yingying Hua and Shiming Ge and Daichi Zhang
Interpret the Predictions of Deep Networks via Re-Label Distillation
by Yingying Hua, Shiming Ge, Daichi Zhang
First submitted to arxiv on: 20 Sep 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
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 This paper proposes a novel method for interpreting the predictions of deep neural networks. The authors develop a re-label distillation approach that learns a direct map from input to prediction in a self-supervision manner. The technique involves projecting images into a VAE subspace, generating synthetic images by perturbing their latent vectors, and annotating these synthetic images based on whether their labels shift. A linear student model is then trained to approximate the annotations, providing a more intuitive explanation for the predictions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In simple terms, this paper helps us understand how deep neural networks make decisions. The authors developed a new way to learn from the network’s own mistakes and provide explanations for its predictions. They did this by creating fake images that are similar to real ones, and then asking the network to label these fake images as “correct” or “incorrect”. By doing so, they can understand how the network makes decisions and provide better explanations. |
Keywords
» Artificial intelligence » Distillation » Student model
