Summary of Robust Yet Efficient Conformal Prediction Sets, by Soroush H. Zargarbashi et al.
Robust Yet Efficient Conformal Prediction Sets
by Soroush H. Zargarbashi, Mohammad Sadegh Akhondzadeh, Aleksandar Bojchevski
First submitted to arxiv on: 12 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- 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 A conformal prediction framework is developed to guarantee that a model’s output includes the true label with any user-specified probability, but it too can be vulnerable to adversarial test examples and perturbed calibration data. To overcome these limitations, robust sets are derived by bounding the worst-case change in conformity scores, leading to more efficient sets for both continuous and discrete data. The guarantees apply to evasion and poisoning attacks on features and labels. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a special kind of prediction that can be trusted to include the correct answer with any level of certainty. The problem is that this type of prediction can also be tricked into giving wrong answers by sneaky test examples or fake calibration data. To solve this, the authors figure out how to limit the worst-case mistake in their predictions, making them more reliable and efficient for both types of data and types of attacks. |
Keywords
* Artificial intelligence * Probability
