Improving group robustness under noisy labels using predictive uncertainty

TL;DR

This paper introduces a novel entropy-based framework that leverages predictive uncertainty to identify spurious-cue-free samples, improving group robustness in noisy label settings.

Contribution

The paper proposes the END framework, which uses predictive uncertainty to identify and oversample SCF samples, enhancing robustness against noisy labels and spurious correlations.

Findings

END outperforms baseline methods on real-world benchmarks.

Predictive uncertainty effectively identifies SCF samples.

Theoretical analysis links high-uncertainty to spurious-cue-free samples.

Abstract

The standard empirical risk minimization (ERM) can underperform on certain minority groups (i.e., waterbirds in lands or landbirds in water) due to the spurious correlation between the input and its label. Several studies have improved the worst-group accuracy by focusing on the high-loss samples. The hypothesis behind this is that such high-loss samples are \textit{spurious-cue-free} (SCF) samples. However, these approaches can be problematic since the high-loss samples may also be samples with noisy labels in the real-world scenarios. To resolve this issue, we utilize the predictive uncertainty of a model to improve the worst-group accuracy under noisy labels. To motivate this, we theoretically show that the high-uncertainty samples are the SCF samples in the binary classification problem. This theoretical result implies that the predictive uncertainty is an adequate indicator to identify SCF samples in a noisy label setting. Motivated from this, we propose a novel ENtropy based Debiasing (END) framework that prevents models from learning the spurious cues while being robust to the noisy labels. In the END framework, we first train the \textit{identification model} to obtain the SCF samples from a training set using its predictive uncertainty. Then, another model is trained on the dataset augmented with an oversampled SCF set. The experimental results show that our END framework outperforms other strong baselines on several real-world benchmarks that consider both the noisy labels and the spurious-cues.