摘要
数据融合算法能够实现对海量数据的整合和特征提取,以便形成更为清晰、可靠的数据,满足不同用户需求,但传统基于BP神经网络的数据融合算存在局部最优及泛化能力差的问题,本文引入了一种无监督学习技术自动编码器,并将其与分簇协议相结合衍生出了新型数据融合算法SAEMAD,最终经过实验对比,在同等条件下,该算法较BPNDA算法具有更好的数据特征提取优势。
data fusion algorithm can realize the integration and feature extraction of the huge data, so as to form a more clear and reliable data, to meet different user requirements, but the traditional data fusion algorithm based on the BP neural network has the problems of local optimum and poor generalization ability, this article introduces a kind of unsupervised learning technology automatic encoder, and combines with clustering protocol to derive an new algorithm ,finally through experiment contrast, under the same condition, the algorithm has better data feature extraction than BPNDA algorithm.
出处
《自动化与仪器仪表》
2017年第9期28-29,34,共3页
Automation & Instrumentation
关键词
深度学习
数据融合技术
层叠自动编码器
分簇协议
deep learning
data fusion technology
cascading automatic encoder
clustering protocols
