Closing the Generalization Gap of Adaptive Gradient Methods in Training Deep Neural Networks

TL;DR

This paper introduces a new adaptive gradient method that combines the fast convergence of Adam with the good generalization of SGD, effectively closing the generalization gap in training deep neural networks.

Contribution

We propose a partially adaptive momentum method that unifies Adam and SGD, with proven convergence and improved generalization performance.

Findings

Our method maintains Adam's fast convergence.

It generalizes as well as SGD in experiments.

It is effective on standard deep learning benchmarks.

Abstract

Adaptive gradient methods, which adopt historical gradient information to automatically adjust the learning rate, despite the nice property of fast convergence, have been observed to generalize worse than stochastic gradient descent (SGD) with momentum in training deep neural networks. This leaves how to close the generalization gap of adaptive gradient methods an open problem. In this work, we show that adaptive gradient methods such as Adam, Amsgrad, are sometimes "over adapted". We design a new algorithm, called Partially adaptive momentum estimation method, which unifies the Adam/Amsgrad with SGD by introducing a partial adaptive parameter $p$, to achieve the best from both worlds. We also prove the convergence rate of our proposed algorithm to a stationary point in the stochastic nonconvex optimization setting. Experiments on standard benchmarks show that our proposed algorithm can maintain a fast convergence rate as Adam/Amsgrad while generalizing as well as SGD in training deep neural networks. These results would suggest practitioners pick up adaptive gradient methods once again for faster training of deep neural networks.