摘要
在基于反向传播神经网络(BPNN)的频谱预测中,网络的初始权重与阈值是随机产生的,并且BP算法本身存在陷入局部最优的缺陷,从而导致BPNN训练得到的网络结构具有一定的不确定性。针对上述问题,提出一种基于HS-BP神经网络的频谱预测算法,通过和声搜索算法的全局寻优能力得到BPNN的最优初始权重和阈值,从而BPNN训练可得到最优的频谱预测网络结构,并运用该网络结构进行频谱的预测。仿真结果表明,该算法可以提高频谱预测的准确性及频谱的利用率。
Initial weight and threshold are random in the process of training Back Propagation Neural Network（ BPNN）. Moreover,BP algorithm is easy to fall into local optimum,so the training of BP neural network is uncertain to get the optimal network structure. To solve this problem,this paper proposes a spectrum prediction method of HS-BP neural network. Using the global optimization capability of harmony search algorithm,it can get the optimal initial weights and thresholds of the BP neural network,ensuring optimal BP neural network structure of training,ultimately,it can use the optimal BP neural network structure of training to predict the spectrum. Simulation results showthat the algorithm can improve the accuracy of spectrum prediction and further improve the utilization of the spectrum.
出处
《计算机工程》
CAS
CSCD
北大核心
2017年第7期146-150,155,共6页
Computer Engineering
基金
国家自然科学基金(71272144)
国家自然科学基金青年基金(61503143)
关键词
认知无线电网络
反向传播神经网络
和声搜索算法
频谱预测
频谱感知
机器学习
Cognitive Radio Network（CRN）
Back Propagation Neural Network（BPNN）
Harmony Search（HS） algorithm
spectrum prediction
spectrum sensing
machine learning
