Summary of Adversarial Training on Purification (atop): Advancing Both Robustness and Generalization, by Guang Lin et al.
Adversarial Training on Purification (AToP): Advancing Both Robustness and Generalization
by Guang Lin, Chao Li, Jianhai Zhang, Toshihisa Tanaka, Qibin Zhao
First submitted to arxiv on: 29 Jan 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 Adversarial Training on Purification (AToP) pipeline combines two components: perturbation destruction by random transforms (RT) and purifier model fine-tuned (FT) by adversarial loss. AToP aims to achieve optimal robustness while generalizing well to unseen attacks, overcoming the limitations of existing methods like Adversarial Training (AT) and Adversarial Purification (AP). By using CIFAR-10, CIFAR-100, and ImageNette datasets, AToP demonstrates its effectiveness in achieving optimal robustness and generalization against unknown attacks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The researchers created a new way to make neural networks more secure. They combined two techniques: one that makes the network forget old tricks and another that helps it learn from mistakes. This combination, called Adversarial Training on Purification (AToP), is better than previous methods at making the network robust against attacks. It works well with different datasets like pictures of animals and objects. |
Keywords
* Artificial intelligence * Generalization
