Accuracy on the Line: on the Strong Correlation Between Out-of-Distribution and In-Distribution Generalization

By John Miller, Rohan Taori, Aditi Raghunathan, Shiori Sagawa, Pang Wei Koh, Vaishaal Shankar, Percy Liang, Yair Carmon, Ludwig Schmidt. 2021

331 citations recorded by the source corpus.

For machine learning systems to be reliable, we must understand their performance in unseen, out-of-distribution environments. In this paper, we empirically show that out-of-distribution performance is strongly correlated with in-distribution performance for a wide range of models and distribution shifts. Specifically, we demonstrate strong correlations between in-distribution and out-of-distribution performance on variants of CIFAR-10&ImageNet, a synthetic pose estimation task derived from YCB objects, satellite imagery classification in FMoW-WILDS, and wildlife classification in iWildCam-WILDS. The strong correlations hold across model architectures, hyperparameters, training set size, and training duration, and are more precise than what is expected from existing domain adaptation theory. To complete the picture, we also investigate cases where the correlation is weaker, for instance some synthetic distribution shifts from CIFAR-10-C and the tissue classification dataset Camelyon17-WILDS. Finally, we provide a candidate theory based on a Gaussian data model that shows how changes in the data covariance arising from distribution shift can affect the observed correlations.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Scaling & Learning Theory, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Peng Cui
2. Renzhe Xu
3. Jindong Wang
4. Xingxu Xie
5. Yue He
6. Hao Chen
7. Yidong Wang
8. Jiashuo Liu

In Computer Vision and Pattern Recognition

1. Peng Cui
2. Renzhe Xu
3. Jindong Wang
4. Yue He
5. Hao Chen
6. Yidong Wang
7. Jiashuo Liu
8. Han Yu

In Safety, Robustness & Interpretability

1. Peng Cui
2. Renzhe Xu
3. Jindong Wang
4. Yue He
5. Jiashuo Liu
6. Han Yu
7. Zheyan Shen
8. Xingxuan Zhang

Citations in the corpus

• Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution (2022, unknown): “We believe the finding that linear probing and LP-FT has higher effective robustness than fine-tuning when the distributon shift is large is particularly interesting because Taori et al. (2020) and Miller et al. (2021) show that it is uncommon for methods to have higher effective robustness.”
• On the Robustness of ChatGPT: An Adversarial and Out-of-distribution Perspective (2023, background): “Finally, our analysis does not show that ID-OOD performances are always positively correlated [29], but can sometimes inversely correlated [47].”
• Towards Out-Of-Distribution Generalization: A Survey (2021, background): “[179] empirically show that the OOD performance is strongly correlated with in-distribution performance (1)https://www.”