论文部分内容阅读
假设检验中Neyman_Person准则是一种基于似然比的信号分类、检测、识别方法 .神经网络是实现这种判定准则的优选方案 ,但是传统的最小平方学习算法 ,如BP算法等 ,往往不能取得全局最优解 .本文针对一种非最小平方学习算法 ,提出了一种概率分配原则 ,并给出了一种Neyman_Person准则的神经网络实现新算法 .文中对新算法在假设检验中的应用进行了仿真验证 ,结果表明新算法具有更小的误差 ,更加适用于Neyman_Person准则 .
Neyman_Person criterion in hypothesis test is a kind of signal classification, detection and recognition method based on likelihood ratio.NN is the best solution to achieve this criterion, but the traditional least squares learning algorithm, such as BP algorithm, often can not get the overall The optimal solution.In this paper, a principle of probability distribution is proposed for a non-least square learning algorithm, and a new algorithm of neural network for Neyman_Person criterion is given. The application of the new algorithm in hypothesis testing is simulated The results show that the new algorithm has smaller error and is more suitable for Neyman_Person criterion.