Switch EMA: A Free Lunch for Better Flatness and Sharpness

TL;DR

Switch EMA (SEMA) is a simple modification to exponential moving average that improves neural network flatness, sharpness, and generalization without extra costs, across diverse tasks and models.

Contribution

The paper introduces SEMA, a straightforward method that switches EMA parameters to the original model after each epoch, enhancing flatness and generalization in DNN training.

Findings

SEMA improves performance across vision and language tasks.

SEMA accelerates convergence speeds.

SEMA enhances the flatness-sharpness trade-off.

Abstract

Exponential Moving Average (EMA) is a widely used weight averaging (WA) regularization to learn flat optima for better generalizations without extra cost in deep neural network (DNN) optimization. Despite achieving better flatness, existing WA methods might fall into worse final performances or require extra test-time computations. This work unveils the full potential of EMA with a single line of modification, i.e., switching the EMA parameters to the original model after each epoch, dubbed as Switch EMA (SEMA). From both theoretical and empirical aspects, we demonstrate that SEMA can help DNNs to reach generalization optima that better trade-off between flatness and sharpness. To verify the effectiveness of SEMA, we conduct comparison experiments with discriminative, generative, and regression tasks on vision and language datasets, including image classification, self-supervised learning, object detection and segmentation, image generation, video prediction, attribute regression, and language modeling. Comprehensive results with popular optimizers and networks show that SEMA is a free lunch for DNN training by improving performances and boosting convergence speeds.