认知网络能够感知外部环境,并能根据周围环境的变化智能、自主、自适应的动态变化,这种特性更适合为用户提供QoS(Quality of Service)保障.设计高精度的流量预测模型,可以提高认知网络的认知特性.本文针对原有预测模型预测精度低、对训练数据依赖程度高以及不能很好的刻画网络流量特征的不足,提出了一个混合的流量预测模型.它使用蚁群算法训练BP网络的权值,避免了梯度下降法收敛速度慢、容易陷入局部最优的问题.并且在预测之前,首先使用BP(Back Propagation)网络剔除原始数据中的异常数据信号,再对其进行小波分解,最后使用混合模型预测网络流量,实现了认知网络中高精度的流量预测.
For a single-relay amplify-and-forward (AF) non-cooperative system,an optimal power proportionbetween source and relay is considered.Aiming to minimize end-to-end bit error rate (BER) and maximizeattainable rate,both large-scale path loss and small-scale Rayleigh fading are taken into account.Aclosed form expression to allocate power in optimal proportion at source is obtained.Simulation resultsshow that the proposed scheme to distribute power can minimize BER under any channel conditions.
