神经网络反传算法改进及其在药物分析中的应用

Modified neural network back propagation algorithm and its application in drug analysis

目的 对神经网络反传算法 (BP)加以改进并用于复方药物的测定。方法 采用改进算法 (MBPN) ,并用XOR问题考察改进前后的收敛速度。结果 将MBPN算法用于复方扑热息痛中三组分的同时测定 ,相对误差为 - 6 .0～ +9.0 %。结论 MBPN法经优化选择传播参数γ加快了神经网络NN的收敛速度 ,增强了反传算法BP的可用性和实用性。用于复方药物的同时测定结果良好、操作简便。

OBJECTIVE To improve the back popagation algorithm and apply it to determin the contents of components in multi-component drug simultaneously. METHOD The MABP has been proposed and checked with XOR to improve the convergent speed before and after improved.RESULTS When the MBPA was applied to the simultaneous determination of the three components of PND, the relative error was between -6.0～+9.0 %.CONCLUSION By selecting optimal parameter of transmission, MBPA can be used to accelerate the convergevent speed of neural network ,develops the usability and practicability of back propagation. When it is applied to determine the contents of the multi-component drug simultaneously, the results show that the method is accurate, and the operation is simple and convenient.