论文部分内容阅读
卷积神经网络(convolutional neural networks,CNN)是深度学习技术应用最成熟的模型之一,利用卷积神经网络来提取特征进行目标识别和分析是当前比较热的研究方向。目前CNN主要以单机串行方式实现,随着大数据时代的到来,串行模式突显出训练时间过长,内存不足等问题。为此,本文提出了一种在分布式处理Hadoop平台上,基于MapReduce框架并行训练CNN的算法MR-TCNN。并通过实验证明,提出的方法与传统单机串行训练方式相比,在大数据上有更快的训练速度。
Convolutional neural networks (CNN) is one of the most mature models of deep learning technology. It is currently hot research direction to use convolutional neural networks to extract features for target recognition and analysis. At present, CNN is mainly implemented in a single serial manner. With the advent of the big data era, the serial mode highlights the problems of long training time and insufficient memory. To this end, this paper presents a distributed processing Hadoop platform, based on MapReduce framework parallel training CNN algorithm MR-TCNN. And through experiments, the proposed method has faster training speed on big data than traditional single-machine serial training method.