Learning Robust Classifiers with Self-Guided Spurious Correlation Mitigation
This addresses the challenge of building robust classifiers without costly spurious correlation annotations, which is important for real-world AI applications, though it is an incremental improvement over existing annotation-free methods.
The paper tackles the problem of deep neural classifiers relying on spurious correlations that harm generalization, by proposing a self-guided framework that automatically constructs fine-grained training labels to mitigate these correlations without needing expensive annotations. The result is improved robustness, outperforming prior methods on five real-world datasets.
Deep neural classifiers tend to rely on spurious correlations between spurious attributes of inputs and targets to make predictions, which could jeopardize their generalization capability. Training classifiers robust to spurious correlations typically relies on annotations of spurious correlations in data, which are often expensive to get. In this paper, we tackle an annotation-free setting and propose a self-guided spurious correlation mitigation framework. Our framework automatically constructs fine-grained training labels tailored for a classifier obtained with empirical risk minimization to improve its robustness against spurious correlations. The fine-grained training labels are formulated with different prediction behaviors of the classifier identified in a novel spuriousness embedding space. We construct the space with automatically detected conceptual attributes and a novel spuriousness metric which measures how likely a class-attribute correlation is exploited for predictions. We demonstrate that training the classifier to distinguish different prediction behaviors reduces its reliance on spurious correlations without knowing them a priori and outperforms prior methods on five real-world datasets.