TL;DR本文提出了一种基于异步随机梯度下降的快速分布式机器学习算法,采用变量规约技术,可使用常量的学习率,并保证线性收敛到最优解,在 Google 云计算平台上的实验表明,该算法在墙时钟时间和解的质量方面优于最先进的分布式异步算法。
Abstract
With the recent proliferation of large-scale learning problems,there have
been a lot of interest on distributed machine learning algorithms, particularly
those that are based on stochastic gradient descent (SGD)