利用分层抽样加速小批量随机梯度下降

May, 2014

Accelerating Minibatch Stochastic Gradient Descent using Stratified Sampling

Peilin Zhao, Tong Zhang

TL;DR本文提出了一种基于分层抽样策略的随机梯度下降算法，可显著改进收敛速度并在实验中得到验证。

Abstract

stochastic gradient descent (SGD) is a popular optimization method which has been applied to many important machine learning tasks such as