When Neural Networks Fail to Generalize? A Model Sensitivity Perspective
This addresses a challenging scenario in domain generalization for machine learning practitioners, though it is incremental as it builds on existing single-DG methods.
The paper tackles the problem of single domain generalization, where only one source domain is available for training, by identifying model sensitivity as a key factor in generalization failure and proposing Spectral Adversarial Data Augmentation (SADA) to suppress this sensitivity, resulting in state-of-the-art performance on multiple datasets.
Domain generalization (DG) aims to train a model to perform well in unseen domains under different distributions. This paper considers a more realistic yet more challenging scenario,namely Single Domain Generalization (Single-DG), where only a single source domain is available for training. To tackle this challenge, we first try to understand when neural networks fail to generalize? We empirically ascertain a property of a model that correlates strongly with its generalization that we coin as "model sensitivity". Based on our analysis, we propose a novel strategy of Spectral Adversarial Data Augmentation (SADA) to generate augmented images targeted at the highly sensitive frequencies. Models trained with these hard-to-learn samples can effectively suppress the sensitivity in the frequency space, which leads to improved generalization performance. Extensive experiments on multiple public datasets demonstrate the superiority of our approach, which surpasses the state-of-the-art single-DG methods.