Summary of Classification-denoising Networks, by Louis Thiry and Florentin Guth
Classification-Denoising Networks
by Louis Thiry, Florentin Guth
First submitted to arxiv on: 4 Oct 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 unified image classification and denoising model is proposed to address the complementary issues of lack of robustness or ignoring conditioning information. The joint probability model combines noisy images and class labels, allowing for both tasks to be performed simultaneously. A forward pass followed by conditioning handles classification, while a denoising function evaluated using the Tweedie-Miyasawa formula computes the score through marginalization and back-propagation. The training objective combines cross-entropy loss with denoising score matching loss integrated over noise levels. Experimental results on CIFAR-10 and ImageNet show competitive performance for both tasks compared to reference models, with improved efficiency and increased robustness to adversarial perturbations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new model helps computers better understand images by combining two important tasks: correctly identifying what’s in the picture and removing noise or distractions. Instead of doing these tasks separately, this model does them together, which makes it more efficient and good at dealing with noisy or misleading information. The results show that this model can accurately identify objects in pictures and remove unwanted noise, and it’s also better at handling fake or “adversarial” images designed to trick computers. |
Keywords
» Artificial intelligence » Classification » Cross entropy » Image classification » Probability
