说明：  BP神经网络算法用来拟合加噪信号，并以动量梯度下降算法和减少内存的Levenberg-Marquardt算法两种训练方法进行该实例下的性能对比
(BP neural network algorithm used to fit the noise signal, and to reduce the momentum gradient algorithm and memory Levenberg-Marquardt algorithm two training methods to compare the performance of the example)