Out-of-distribution Generalization in the Presence of Nuisance-Induced Spurious Correlations

TL;DR

This paper introduces Nuisance-Randomized Distillation (NURD), a method for training models that maintain high performance across varying nuisance-label relationships by focusing on representations independent of nuisances.

Contribution

The paper proposes a novel approach, NURD, that finds representations invariant to nuisances, ensuring robust out-of-distribution generalization in the presence of spurious correlations.

Findings

NURD outperforms baseline models in chest X-ray classification with spurious background correlations.

NURD's representations are independent of nuisance variables under the nuisance-randomized distribution.

Models trained with NURD maintain high accuracy across different nuisance-label relationships.

Abstract

In many prediction problems, spurious correlations are induced by a changing relationship between the label and a nuisance variable that is also correlated with the covariates. For example, in classifying animals in natural images, the background, which is a nuisance, can predict the type of animal. This nuisance-label relationship does not always hold, and the performance of a model trained under one such relationship may be poor on data with a different nuisance-label relationship. To build predictive models that perform well regardless of the nuisance-label relationship, we develop Nuisance-Randomized Distillation (NURD). We introduce the nuisance-randomized distribution, a distribution where the nuisance and the label are independent. Under this distribution, we define the set of representations such that conditioning on any member, the nuisance and the label remain independent. We prove that the representations in this set always perform better than chance, while representations outside of this set may not. NURD finds a representation from this set that is most informative of the label under the nuisance-randomized distribution, and we prove that this representation achieves the highest performance regardless of the nuisance-label relationship. We evaluate NURD on several tasks including chest X-ray classification where, using non-lung patches as the nuisance, NURD produces models that predict pneumonia under strong spurious correlations.