Improving Layer-wise Adaptive Rate Methods using Trust Ratio Clipping

TL;DR

This paper introduces LAMBC, a variant of LAMB that uses trust ratio clipping to stabilize training, improving large batch neural network training accuracy and stability.

Contribution

The paper proposes trust ratio clipping in LAMB to address instability issues, enhancing large batch training performance.

Findings

LAMBC improves training stability across batch sizes.

Empirical results show better accuracy on ImageNet and CIFAR-10.

Trust ratio clipping reduces extreme trust ratio values.

Abstract

Training neural networks with large batch is of fundamental significance to deep learning. Large batch training remarkably reduces the amount of training time but has difficulties in maintaining accuracy. Recent works have put forward optimization methods such as LARS and LAMB to tackle this issue through adaptive layer-wise optimization using trust ratios. Though prevailing, such methods are observed to still suffer from unstable and extreme trust ratios which degrades performance. In this paper, we propose a new variant of LAMB, called LAMBC, which employs trust ratio clipping to stabilize its magnitude and prevent extreme values. We conducted experiments on image classification tasks such as ImageNet and CIFAR-10 and our empirical results demonstrate promising improvements across different batch sizes.