Calibrated ensembles can mitigate accuracy tradeoffs under distribution shift
By Ananya Kumar, Tengyu Ma, Percy Liang, Aditi Raghunathan. 2022
48 citations recorded by the source corpus.
We often see undesirable tradeoffs in robust machine learning where out-of-distribution (OOD) accuracy is at odds with in-distribution (ID) accuracy: a robust classifier obtained via specialized techniques such as removing spurious features often has better OOD but worse ID accuracy compared to a standard classifier trained via ERM. In this paper, we find that ID-calibrated ensembles -- where we simply ensemble the standard and robust models after calibrating on only ID data -- outperforms prior state-of-the-art (based on self-training) on both ID and OOD accuracy. On eleven natural distribution shift datasets, ID-calibrated ensembles obtain the best of both worlds: strong ID accuracy and OOD accuracy. We analyze this method in stylized settings, and identify two important conditions for ensembles to perform well both ID and OOD: (1) we need to calibrate the standard and robust models (on ID data, because OOD data is unavailable), (2) OOD has no anticorrelated spurious features.
Subfields: Artificial Intelligence, 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
- Mohammad Rastegari
- Sachin Mehta
- Oncel Tuzel
- Hadi Pouransari
- Ali Farhadi
- Mehrdad Farajtabar
- Fartash Faghri
- Dacheng Tao
In Safety, Robustness & Interpretability
Citations in the corpus
- Reinforce Data, Multiply Impact: Improved Model Accuracy and Robustness with Dataset Reinforcement (2023, background): “This matches recent observations about ensembles that out-of-distribution robustness is better for well-calibrated models [33].”