Navigate Beyond Shortcuts: Debiased Learning through the Lens of Neural Collapse

TL;DR

This paper investigates Neural Collapse in neural networks trained on biased datasets, identifying how shortcut learning hampers generalization, and proposes a prime training framework to promote unbiased feature learning, achieving state-of-the-art results.

Contribution

It introduces a shortcut prime training method based on Neural Collapse to mitigate bias and improve generalization without extra training complexity.

Findings

Models avoid shortcuts with prime training.

Enhanced convergence during training.

State-of-the-art performance on biased datasets.

Abstract

Recent studies have noted an intriguing phenomenon termed Neural Collapse, that is, when the neural networks establish the right correlation between feature spaces and the training targets, their last-layer features, together with the classifier weights, will collapse into a stable and symmetric structure. In this paper, we extend the investigation of Neural Collapse to the biased datasets with imbalanced attributes. We observe that models will easily fall into the pitfall of shortcut learning and form a biased, non-collapsed feature space at the early period of training, which is hard to reverse and limits the generalization capability. To tackle the root cause of biased classification, we follow the recent inspiration of prime training, and propose an avoid-shortcut learning framework without additional training complexity. With well-designed shortcut primes based on Neural Collapse structure, the models are encouraged to skip the pursuit of simple shortcuts and naturally capture the intrinsic correlations. Experimental results demonstrate that our method induces better convergence properties during training, and achieves state-of-the-art generalization performance on both synthetic and real-world biased datasets. Our code is available at https://github.com/RachelWolowitz/Navigate-beyond-Shortcuts/tree/main.