论文部分内容阅读
目的:利用自组织竞争型神经网络判别不同产地青果的指纹图谱,为青果的质量评价奠定基础。方法:色谱条件采用Phenomenex Luna C18色谱柱(2)100A(4.6 mm×250 mm,5μm),流动相乙腈-1%甲酸,检测波长270 nm,流速0.8 m L·min~(-1),柱温20℃,建立不同产地青果的指纹图谱,利用竞争层神经元数目为3,学习率为0.01,收敛次数为690的自组织竞争型人工神经网络模型对其进行分类判别。结果:自组织竞争型神经网络模型对青果HPLC指纹图谱分类平均错误率为39.13%。结论:自组织竞争型神经网络模型无法将不同产地的青果有效分类,不同产地的青果化学成分种类及含量差异不明显。
OBJECTIVE: To identify the fingerprints of young fruits from different regions by self-organizing competitive neural network, and lay the foundation for the quality evaluation of young fruits. METHODS: The chromatographic conditions were as follows: 100 (4.6 mm × 250 mm, 5 μm) on a Phenomenex Luna C18 column with a mobile phase of acetonitrile-1% formic acid at a detection wavelength of 270 nm and a flow rate of 0.8 mL · min -1 The temperature was 20 ℃, the fingerprints of green fruit from different areas were established. The self-organizing competitive artificial neural network model was used to discriminate the fingerprints of juvenile fruit of different origins by using the number of neurons in competition layer 3, the learning rate of 0.01 and the number of convergences of 690. Results: The average error rate of self-organization competitive neural network model in classification of HPLC fingerprint was 39.13%. CONCLUSION: Self-organizing competitive neural network model can not effectively classify young fruits from different habitats. There is no obvious difference in the types and contents of the chemical components in different producing areas.