Summary of Generalization Vs. Specialization Under Concept Shift, by Alex Nguyen et al.
Generalization vs. Specialization under Concept Shift
by Alex Nguyen, David J. Schwab, Vudtiwat Ngampruetikorn
First submitted to arxiv on: 23 Sep 2024
Categories
- Main: Machine Learning (stat.ML)
- Secondary: Disordered Systems and Neural Networks (cond-mat.dis-nn); Statistical Mechanics (cond-mat.stat-mech); 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 Machine learning models are often brittle under distribution shift, i.e., when data distributions at test time differ from those during training. The paper analyzes ridge regression under concept shift, a type of distribution shift where the input-label relationship changes at test time. It derives an exact expression for prediction risk in the high-dimensional limit and reveals nontrivial effects on generalization performance depending on robust and nonrobust features. The results show that test performance can exhibit nonmonotonic data dependence, even when double descent is absent. Experiments on MNIST and FashionMNIST suggest this behavior also occurs in classification problems. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine learning models often struggle when the rules change at test time. This paper looks at how ridge regression performs when the relationship between inputs and labels changes. It shows that performance can be affected by different types of features, and that it’s possible to see a non-linear effect even if everything seems okay. The results are important for understanding how machine learning models behave in real-world situations. |
Keywords
» Artificial intelligence » Classification » Generalization » Machine learning » Regression
