【摘要】 为了扩大邻域函数的输出空间和增强神经元的邻域合作,提出基于q-高斯的SOM(self-organizing mapping)神经网络评估雷达抗干扰效能.采用q-高斯函数作为SOM神经网络的邻域函数,选取较大的非广延熵指数q扩大了q-高斯函数的输出空间,随着邻域的缩小,非广延熵指数q从大到小自适应地调整平衡了神经元的远邻域合作和近邻域合作.通过评估雷达抗干扰效能和实例测试,仿真结果表明基于q-高斯的SOM神经网络效能评估的准确率为100%、模式识别的聚类正确率和分类正确率比其他SOM神经网络高出5%,验证了该方法的有效性和可行性.更多还原

【Abstract】 In order to increase the output space of neighborhood functions and enhance the neighborhood cooperation between neurons,a q-Gaussian self-organizing mapping(SOM) neural network was proposed for evaluation of the effectiveness of radar electronic counter-countermeasures(ECCM).A q-Gaussian function was taken as a neighborhood function in an SOM neural network,and the non-extensive entropic index q was larger to efficiently increase the output space of the q-Gaussian function.The non-extensive entropic index q was adjusted adaptively from large to small with the decreasing neighbor to balance the neurons’ distant and close neighborhood cooperation ability.The simulation results of the effectiveness evaluation of the radar ECCM and instance tests show that the q-Gaussian SOM neural network can obtain 100% accurate results in evaluating effectiveness,a 5% higher accuracy rate both in clustering and classification than other SOM neural networks in pattern recognition;the validity and feasibility of the method are verified.更多还原