Spuriosity Didn't Kill the Classifier: Using Invariant Predictions to Harness Spurious Features
This addresses the challenge of using unstable features in machine learning to boost performance on out-of-distribution data, offering a novel method for domain adaptation.
The paper tackles the problem of leveraging spurious features to improve out-of-distribution performance without test-domain labels, showing that Stable Feature Boosting (SFB) can learn an asymptotically-optimal predictor and demonstrating its effectiveness on real and synthetic data.
To avoid failures on out-of-distribution data, recent works have sought to extract features that have an invariant or stable relationship with the label across domains, discarding "spurious" or unstable features whose relationship with the label changes across domains. However, unstable features often carry complementary information that could boost performance if used correctly in the test domain. In this work, we show how this can be done without test-domain labels. In particular, we prove that pseudo-labels based on stable features provide sufficient guidance for doing so, provided that stable and unstable features are conditionally independent given the label. Based on this theoretical insight, we propose Stable Feature Boosting (SFB), an algorithm for: (i) learning a predictor that separates stable and conditionally-independent unstable features; and (ii) using the stable-feature predictions to adapt the unstable-feature predictions in the test domain. Theoretically, we prove that SFB can learn an asymptotically-optimal predictor without test-domain labels. Empirically, we demonstrate the effectiveness of SFB on real and synthetic data.