摘要
基于雷达信号分选的应用需求,对传统自组织特征映射(SOFM)神经网络进行了分析,提出一种结构自调整的改进型SOFM神经网络,并进行了分选仿真.结果表明,该网络能够通过增加或删减输出神经元的数目,自动调整神经网络的规模使其与实际输入模式的类别数目相适应,从而取得了较好的雷达信号分选效果.
Traditional self-organizing feature mapping (SOFM) neural network is analysised based on the application requirements for radar signal sorting. An improved SOFM neural network with self-adjusting is presented and sorting simulation is carried out. According to the results of simulation, the network can adaptively adjust its scope by increasing or decreasing the number of output neuron, to accommodate the number of actual input pattern sorting and gain the better performance for sorting radar signal.
出处
《空军雷达学院学报》
2005年第2期18-20,共3页
Journal of Air Force Radar Academy
关键词
SOFM神经网络
结构自调整
雷达信号分选
self-organizing feature map (SOFM) neural network
structure self-adjusting
radar signal sorting