Towards Understanding Label Smoothing

TL;DR

This paper provides a theoretical analysis of label smoothing regularization (LSR) in deep neural network training, showing how it accelerates convergence by reducing variance and proposing a two-stage LSR strategy that improves training efficiency.

Contribution

It offers the first convergence analysis of LSR in non-convex optimization and introduces a two-stage LSR method that enhances training speed and effectiveness.

Findings

LSR reduces variance and speeds up convergence in stochastic gradient descent.

Two-stage LSR (TSLA) improves training efficiency by applying LSR early and dropping it later.

Empirical results show TSLA outperforms baseline methods on ResNet training.

Abstract

Label smoothing regularization (LSR) has a great success in training deep neural networks by stochastic algorithms such as stochastic gradient descent and its variants. However, the theoretical understanding of its power from the view of optimization is still rare. This study opens the door to a deep understanding of LSR by initiating the analysis. In this paper, we analyze the convergence behaviors of stochastic gradient descent with label smoothing regularization for solving non-convex problems and show that an appropriate LSR can help to speed up the convergence by reducing the variance. More interestingly, we proposed a simple yet effective strategy, namely Two-Stage LAbel smoothing algorithm (TSLA), that uses LSR in the early training epochs and drops it off in the later training epochs. We observe from the improved convergence result of TSLA that it benefits from LSR in the first stage and essentially converges faster in the second stage. To the best of our knowledge, this is the first work for understanding the power of LSR via establishing convergence complexity of stochastic methods with LSR in non-convex optimization. We empirically demonstrate the effectiveness of the proposed method in comparison with baselines on training ResNet models over benchmark data sets.