论文部分内容阅读
将以误差反向传播为训练算法的前馈式人工神经网络（BP-ANN）首次用于中草药的裂解气相色谱谱图解析。重点考察了如何表征和提取复杂的裂解色谱图中有价值信息，用主成分分析方法处理后输入到参数经优化的神经网络中。实验证明，该方法不仅可以正确识别样品所属种类，而且对于不同实验时间、数据残缺等原因造成的噪音具有优异的抗干扰能力。
For the first time, the feedforward artificial neural network (BP-ANN), which propagates the error backwards as a training algorithm, is used for the analysis of Chinese herbal medicines by pyrolysis gas chromatography. We focus on how to characterize and extract valuable information from complex chromatograms, process it with principal component analysis, and input it into the optimized neural network. Experiments show that this method not only can correctly identify the type of sample, but also has excellent anti-interference ability against noise caused by different experimental time, data incompleteness and so on.