Summary of Robust Nas Under Adversarial Training: Benchmark, Theory, and Beyond, by Yongtao Wu et al.
Robust NAS under adversarial training: benchmark, theory, and beyond
by Yongtao Wu, Fanghui Liu, Carl-Johann Simon-Gabriel, Grigorios G Chrysos, Volkan Cevher
First submitted to arxiv on: 19 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (stat.ML)
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 neural architecture search (NAS) framework is introduced to address the lack of benchmark evaluations and theoretical guarantees for searching robust architectures against malicious data. The authors release a comprehensive dataset that includes both clean accuracy and robust accuracy for various adversarially trained networks from the NAS-Bench-201 search space on image datasets. Additionally, they establish a generalization theory for searching architectures in terms of clean accuracy and robust accuracy under multi-objective adversarial training using the neural tangent kernel (NTK) tool. This work is expected to benefit the NAS community by providing reliable reproducibility, efficient assessment, and theoretical foundation for the pursuit of robust architectures. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps researchers find better ways to design artificial intelligence models that can withstand attacks from malicious data. The authors create a big dataset of different AI models and how well they perform on clean and dirty data. They also figure out why some AI models are better than others at handling dirty data. This work will make it easier for other researchers to develop new AI models that are more robust against attacks. |
Keywords
* Artificial intelligence * Generalization
