# Preventing Failures Due to Dataset Shift: Learning Predictive Models That Transport

arXiv: https://arxiv.org/abs/1812.04597 · HTML: https://mlsanity.com/paper/1812.04597/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 11 Dec 2018.

## Built on by (most cited first, 6 of 6)

- [Algorithm Fairness in AI for Medicine and Healthcare](https://mlsanity.com/paper/2110.00603/) (2021)
- [No Fair Lunch: A Causal Perspective on Dataset Bias in Machine Learning for Medical Imaging](https://mlsanity.com/paper/2307.16526/) (2023)
- [Do causal predictors generalize better to new domains?](https://mlsanity.com/paper/2402.09891/) (2024)
- [Improving generalization of machine learning-identified biomarkers with causal modeling: an investigation into immune receptor diagnostics](https://mlsanity.com/paper/2204.09291/) (2022)
- [Effect-Invariant Mechanisms for Policy Generalization](https://mlsanity.com/paper/2306.10983/) (2023)
- [Implicit Design Choices and Their Impact on Emotion Recognition Model Development and Evaluation](https://mlsanity.com/paper/2309.03238/) (2023)

## Abstract (arXiv)

Classical supervised learning produces unreliable models when training and target distributions differ, with most existing solutions requiring samples from the target domain. We propose a proactive approach which learns a relationship in the training domain that will generalize to the target domain by incorporating prior knowledge of aspects of the data generating process that are expected to differ as expressed in a causal selection diagram. Specifically, we remove variables generated by unstable mechanisms from the joint factorization to yield the Surgery Estimator---an interventional distribution that is invariant to the differences across environments. We prove that the surgery estimator finds stable relationships in strictly more scenarios than previous approaches which only consider conditional relationships, and demonstrate this in simulated experiments. We also evaluate on real world data for which the true causal diagram is unknown, performing competitively against entirely data-driven approaches.