Identifying Spurious Biases Early in Training through the Lens of Simplicity Bias

TL;DR

This paper provides a theoretical analysis of how neural networks' preference for simpler solutions causes early learning of spurious correlations, and introduces SPARE, a method to detect and mitigate these biases effectively.

Contribution

The paper offers the first theoretical insights into simplicity bias's role in learning spurious correlations and proposes SPARE, a fast, effective method for early detection and mitigation.

Findings

SPARE outperforms state-of-the-art in worst-group accuracy by up to 21.1%.

SPARE is up to 12 times faster than existing methods.

Early in training, models' outputs reveal spurious features with high confidence.

Abstract

Neural networks trained with (stochastic) gradient descent have an inductive bias towards learning simpler solutions. This makes them highly prone to learning spurious correlations in the training data, that may not hold at test time. In this work, we provide the first theoretical analysis of the effect of simplicity bias on learning spurious correlations. Notably, we show that examples with spurious features are provably separable based on the model's output early in training. We further illustrate that if spurious features have a small enough noise-to-signal ratio, the network's output on the majority of examples is almost exclusively determined by the spurious features, leading to poor worst-group test accuracy. Finally, we propose SPARE, which identifies spurious correlations early in training and utilizes importance sampling to alleviate their effect. Empirically, we demonstrate that SPARE outperforms state-of-the-art methods by up to 21.1% in worst-group accuracy, while being up to 12x faster. We also show that SPARE is a highly effective but lightweight method to discover spurious correlations.