论文部分内容阅读
针对认知无线网络中基于神经网络的频谱预测模型训练低效的问题,提出一种基于“模更新”的高效训练方法.基于“模更新”的训练方法通过在判决前对神经网络的预测输出和期望输出计算预测误差,使频谱预测模型可以迭代更新神经网络,在预测性能不降低的前提下提高了频谱预测模型的训练效率.根据“模更新”训练得到的预测误差定义预测准确度估计变量,作为评估频谱预测是否准确的依据,选择预测准确度估计值高的信道作为感知信道,降低了选择发生频谱预测错误信道的概率,仿真结果表明该方法提高了认知无线网络中次级用户的发送概率和平均吞吐率.“模更新”训练方法不仅提高了频谱预测模型的训练效率,还提高了认知无线网络中次级用户的性能.
In order to solve the inefficient training problem of neural network-based spectrum prediction model in cognitive wireless networks, an efficient training method based on “modal updating” is proposed. The training method based on “modulo updating” And predicting the output prediction error so that the spectrum prediction model can iteratively update the neural network to improve the training efficiency of the spectrum prediction model without reducing the prediction performance.Depending on the prediction error obtained from the “module updating” training, the estimation accuracy of the prediction accuracy As a basis for assessing the accuracy of spectrum prediction, the channel with high prediction accuracy is selected as the perceptual channel, and the probability of selecting the channel for error prediction is reduced. Simulation results show that this method can improve the accuracy of sub-users Sending probability and average throughput.The “modulo update” training method not only improves the training efficiency of the spectrum predictive model, but also improves the secondary user’s performance in the cognitive wireless network.