MaxVA: Fast Adaptation of Step Sizes by Maximizing Observed Variance of Gradients

TL;DR

MaxVA introduces a novel adaptive learning rate method that maximizes the observed variance of gradients, leading to faster adaptation and improved convergence in training deep models like Transformers.

Contribution

It replaces the running mean in Adam with a weighted mean optimized to maximize gradient variance, enhancing adaptation and convergence.

Findings

Faster convergence in training deep models.

Improved empirical performance over Adam.

Effective in large-batch pretraining of BERT.

Abstract

Adaptive gradient methods such as RMSProp and Adam use exponential moving estimate of the squared gradient to compute adaptive step sizes, achieving better convergence than SGD in face of noisy objectives. However, Adam can have undesirable convergence behaviors due to unstable or extreme adaptive learning rates. Methods such as AMSGrad and AdaBound have been proposed to stabilize the adaptive learning rates of Adam in the later stage of training, but they do not outperform Adam in some practical tasks such as training Transformers \cite{transformer}. In this paper, we propose an adaptive learning rate principle, in which the running mean of squared gradient in Adam is replaced by a weighted mean, with weights chosen to maximize the estimated variance of each coordinate. This results in a faster adaptation to the local gradient variance, which leads to more desirable empirical convergence behaviors than Adam. We prove the proposed algorithm converges under mild assumptions for nonconvex stochastic optimization problems, and demonstrate the improved efficacy of our adaptive averaging approach on machine translation, natural language understanding and large-batch pretraining of BERT. The code is available at https://github.com/zhuchen03/MaxVA.