改进的BP网络用于气相色谱保留指数预测

改进了ＢＰ神经网络的学习速率算法，并运用改进后的ＢＰ网络，通过分子结构参数，同时对连载在吡啶在非极性固定液ＳＥ－３０和极性固定液ＰＥＧ－１５００柱上的保留指数进行预测，取得了较好的结果。

Using Improved BP Neural Network in Predicting GC Retention Indices

The back-propagation neural network model is improved so that its learning rate can be automatically changed with the learning error. Based on the improved BP model, the relationship between the molecular parameters of a set of alkylpyridine and their GC retention indices observed on both SE-30 and PEG-1 500 columns is studied. The relative mean error (RME) of prediction is lower than 2%. The result also shows that the performance of improved BP model is better than the original one.