Summary of On Temperature Scaling and Conformal Prediction Of Deep Classifiers, by Lahav Dabah et al.
On Temperature Scaling and Conformal Prediction of Deep Classifiers
by Lahav Dabah, Tom Tirer
First submitted to arxiv on: 8 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: 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 This paper investigates the interplay between classification confidence indicators in deep neural networks (DNNs). Specifically, it explores the relationship between Temperature Scaling (TS) calibration and Conformal Prediction (CP) methods. The authors demonstrate that TS calibration improves class-conditional coverage of adaptive CP methods but negatively affects prediction set sizes. They then develop a mathematical theory to explain this non-monotonic trend and provide guidelines for practitioners to effectively combine adaptive CP with calibration. The paper’s findings have implications for designing accurate and reliable classification systems in various applications. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how well computers can predict things. It tries to figure out how different ways of telling a computer how confident it should be affect its ability to make good predictions. The researchers found that making the computer more confident actually makes it worse at predicting which option is most likely. They also developed some math to explain why this happens and gave advice on how to use this information to make better computer programs. |
Keywords
* Artificial intelligence * Classification * Temperature
