摘要
将基于遗传算法(Genetic algorithm,GA)用于优化径向基函数(Radial Basis Function,RBF)神经网络的输入变量,以提高RBF神经网络的定量分析重叠的同步荧光光谱的能力。本文提出的基于GA输入变量选择的RBF神经网络可作为一种消除光谱干扰的有效工具。光谱对应的有关数据可作为RBF神经网的输入变量,这些多元变量使得神经网络在训练过程中产生"过拟合"现象,降低了定量分析的准确度。用GA优化RBF神经网的输入变量,既简化了神经网络的结构又提高了神经网络的学习能力。通过分析模拟数据和实验数据的计算结果,该方法用于提高RBF人工神经网络网的学习能力可行,且有效。
The genetic algorithm （GA） was used in optimizing input variables of the artificial neural network （ANN） to improve the precision of quantitative analysis of the unresolved synchronous fluorescence spectra by radial basis function （RBF）-ANN. The proposed method is a powerful tool to resolve the spectra interference. The data corresponding spectra of selected wavelengths can be used as the input data of ANN, but the multivariate may cause ＂over-fitting＂ of the trained networks, which could decrease the precision of quantification. The strategy based on GA make input variables of ANN be optimized, the model be simplified and learning ability of ANN be improved to some extent. Through analyzing the calculation of the simulation data and the experimental data, the method is practical and effective in optimizing the RBF-ANN.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2013年第2期171-175,共5页
Computers and Applied Chemistry
基金
河北省自然科学基金资助项目(B2008000583)
关键词
径向基函数神经网络
定量分析
同步荧光
优化
遗传算法
radial basis function, quantitation, synchronous fluorescence, optimization, genetic algorithm
