用Matlab编程，实现解决该问题的单样本训练BP网络，设置一个停止迭代的误差Emin和最大迭代次数。在调试过程中，通过不断调整隐层节点数，学习率η，找到收敛速度快且误差小的一组参数。产生均匀分布在区间[-4，4]的测试样本，输入建立的模型得到输出，与Hermit多项式的期望输出进行比较计算总误差（运行5次，取平均值）-Using Matlab programming, to solve the problem of single sample training BP network, set a stop error iterations Emin and the maximum number of iterations. In the debugging process, by continuously adjusting the hidden layer nodes, learning rate η, fast convergence and find a set of parameters small errors. Produce a uniform distribution in the interval [4,4] of the test sample, the model has been established input output with the expected output Hermit polynomial calculated by comparing the total error (run 5 times the average)