Learning Robust Classifiers with Self-Guided Spurious Correlation Mitigation

TL;DR

This paper introduces a self-guided framework that automatically mitigates spurious correlations in deep classifiers without requiring annotated data, enhancing robustness by leveraging a novel spuriousness embedding space.

Contribution

It proposes an annotation-free method that constructs fine-grained training labels using a new spuriousness metric and embedding space to improve classifier robustness.

Findings

Outperforms prior methods on five real-world datasets.

Reduces reliance on spurious correlations without prior annotations.

Automatically detects conceptual attributes for training label refinement.

Abstract

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.