摘要
将粗糙神经元和传统神经元混合构造的粗糙神经网络,用于对毫米波辐射计回波信号的目标识别。其中粗糙神经元包含一对重叠的普通神经元,使用一对上下值作为输入和输出。对于实际应用中变量值是范围值的情况,用粗糙神经网络来开发模型,结果优于传统神经网络。仿真实验表明,该模型提高了目标的识别率和网络的收敛速度。
In this paper,a rough neural network composed of rough neuron and traditional neuron is used to target recognition of millimeter wave radiometer signal.Rough neuron contains a pair of overlapped common neurons,using a pair of values from top to bottom as input and output.For the variable value being scope value in practical application,rough neural network is used to develop the model,so that better results can be achieved.Simulation results show that this model improves the rate of target recognition and the convergence speed of network.
出处
《探测与控制学报》
CSCD
北大核心
2008年第5期38-40,共3页
Journal of Detection & Control
关键词
神经网络
粗集理论
粗糙神经元
目标识别
neural networks
rough set theory
rough neuron
target identification