Why Does Sharpness-Aware Minimization Generalize Better Than SGD?
This provides insight into a specific training method for neural networks, addressing overfitting in classification tasks, but it is incremental as it builds on existing SAM explanations.
The paper tackled the problem of understanding why Sharpness-Aware Minimization (SAM) generalizes better than Stochastic Gradient Descent (SGD) in neural networks, demonstrating that SAM prevents noise learning early on and facilitates feature learning, with experiments on synthetic and real data supporting the theory.
The challenge of overfitting, in which the model memorizes the training data and fails to generalize to test data, has become increasingly significant in the training of large neural networks. To tackle this challenge, Sharpness-Aware Minimization (SAM) has emerged as a promising training method, which can improve the generalization of neural networks even in the presence of label noise. However, a deep understanding of how SAM works, especially in the setting of nonlinear neural networks and classification tasks, remains largely missing. This paper fills this gap by demonstrating why SAM generalizes better than Stochastic Gradient Descent (SGD) for a certain data model and two-layer convolutional ReLU networks. The loss landscape of our studied problem is nonsmooth, thus current explanations for the success of SAM based on the Hessian information are insufficient. Our result explains the benefits of SAM, particularly its ability to prevent noise learning in the early stages, thereby facilitating more effective learning of features. Experiments on both synthetic and real data corroborate our theory.