Oral
in
Workshop: Workshop on Spurious Correlation and Shortcut Learning: Foundations and Solutions
Mitigating Simplicity Bias in Neural Networks: A Feature Sieve Modification, Regularization, and Self-Supervised Augmentation Approach
Gaurav Joshi · Parth Shah · Rachit Verma
Keywords: [ Regularization ] [ Simplicity bias ] [ Feature Sieve ] [ Self-Supervised Learning ] [ Neuronal Correlation ]
Neural networks (NNs) are known to exhibit simplicity bias, where they tend to prioritize learning simple features over more complex ones, even when the latter are more informative. This bias can result in models making skewed predictions with poor out-of-distribution (OOD) generalization. To address this issue, we propose three techniques to mitigate simplicity bias. One of these is a modification to the Feature Sieve method. In the second method we utilize neuronal correlations as a penalizing effect to try and enforce the learning of different features. The third technique involves a novel feature-building approach called Self-Supervised Augmentation. We validate our methods' generalization capabilities through experiments on a custom dataset.