摘要
针对传统发射区解算多项式拟合方法存在多项式函数难确定、函数分段范围难把握的问题,采用了BP神经网络(BPNN)实现数据整体拟合。当数据量大且复杂时,拓展神经网络的深度不能完整学习数据特征。针对该问题,引入了深度置信网(DBN)。在DBN仿真训练中,选用双隐层网络结构,随着隐单元数的增加,拟合性能有较明显提升。仿真结果表明,深度学习方法更适用于大数据环境下的深度神经网络架构,可应用于地空导弹发射区解算。
Aimed at the problem of the traditional polynomial fitting method for launch area about difficult to confirm the polynomial function and difficult to hold the function piecewise range,the back propagation neural network(BPNN)is adopted to realize fitting for the whole data.When facing the huge date amount and the complex data,the data characteristics can't be perfectly learned by expanding the depth of the neural network.Aimed at the problem of the BPNN,the deep belief net(DBN)is introduced.During the simulated training of the DBN,the network structure with double hidden layers is adopted.With the addition of the hidden units,the fitting performance is increased.The simulation result shows that the deep learning method is suitable for the DBN architecture in the big data environment and it can be applied in calculating the launch area of surface to air missile(SAM).
作者
薛亚勇
胡国文
XUE Yayong;HU Guowen(The 28th Research Institute of China Electronics Technology Group Corporation,Nanjing 210007,China)
出处
《指挥信息系统与技术》
2018年第4期48-52,共5页
Command Information System and Technology
关键词
地空导弹发射区
BP神经网络
深度学习
深度置信网
launch area of surface to air missile (SAM)
back propagation neural network(BPNN)
deep learning
deep belief net (DBN)