On Distributed Deep Network for Processing Large-Scale Sets of Complex Data
Qin Chao, Xiaoguang Gao, Chen Da-Qing · 2016
Recent work in unsupervised feature learning and deep learning has shown that being able to train large models can dramatically improve performance. In this paper, we consider the problem of training a deep network with hundreds of parameters using distributed CPU cores. We have developed Bagging-Down SGD algorithm to solve the distributing problems. Bagging-Down SGD introduces the parameter server adding on the several model replicas, and separates the updating and the training computing to accelerate the whole system. We have successfully used our system to train a distributed deep network, and achieve stateof-the-art performance on MINIST, a visual handwriting font library. We show that these techniques dramatically accelerate the training of this kind of distributed deep network.