# Robust Prediction when Features are Missing

arXiv: https://arxiv.org/abs/1912.07226 · HTML: https://mlsanity.com/paper/1912.07226/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 16 Dec 2019.

## Built on by (most cited first, 5 of 5)

- [Sharing pattern submodels for prediction with missing values](https://mlsanity.com/paper/2206.11161/) (2022)
- [Estimation under Model Misspecification with Fake Features](https://mlsanity.com/paper/2203.03398/) (2022)
- [Model Mismatch Trade-offs in LMMSE Estimation](https://mlsanity.com/paper/2105.11964/) (2021)
- [Regularization Trade-offs with Fake Features](https://mlsanity.com/paper/2212.00433/) (2022)
- [Adversarial Training of Linear Models under Stealthy Attacks](https://mlsanity.com/paper/2608.25681/) (2026)

## Abstract (arXiv)

Predictors are learned using past training data which may contain features that are unavailable at the time of prediction. We develop an approach that is robust against outlying missing features, based on the optimality properties of an oracle predictor which observes them. The robustness properties of the approach are demonstrated on both real and synthetic data.