On the Generalization Benefit of Noise in Stochastic Gradient Descent
This work addresses a foundational problem in machine learning optimization, confirming that noise in SGD enhances generalization, which is incremental as it resolves recent controversies but builds on long-standing arguments.
The paper tackles the debate over whether minibatch stochastic gradient descent (SGD) generalizes better than large-batch gradient descent in deep neural networks, finding through experiments that small or moderately large batch sizes substantially outperform very large batches on test sets, even with equal training iterations and lower training losses for large batches.
It has long been argued that minibatch stochastic gradient descent can generalize better than large batch gradient descent in deep neural networks. However recent papers have questioned this claim, arguing that this effect is simply a consequence of suboptimal hyperparameter tuning or insufficient compute budgets when the batch size is large. In this paper, we perform carefully designed experiments and rigorous hyperparameter sweeps on a range of popular models, which verify that small or moderately large batch sizes can substantially outperform very large batches on the test set. This occurs even when both models are trained for the same number of iterations and large batches achieve smaller training losses. Our results confirm that the noise in stochastic gradients can enhance generalization. We study how the optimal learning rate schedule changes as the epoch budget grows, and we provide a theoretical account of our observations based on the stochastic differential equation perspective of SGD dynamics.