论文部分内容阅读
多元分析与滤波校正已广泛用于多组分同时或选择测定,但其中许多方法均基於线性加和性原理,而此定量原理受各种理化因素制约并非总能满足.神经网络NN如反传算法BP则处理方式完全不同,毋需知道或采用任何形式的输入/输出关系模型,它依据一定学习规则处理问题,特别对因果关系不明确、知识背景不清楚、推量规则不确定的问题求解具独到之处.我们曾用NN研究定量构效关系、化学模式识别、生化反应建模、复杂机理剖析及多元光谱分析等,
Multivariate analysis and filter correction have been widely used for the simultaneous or selective determination of multiple components, but many of them are based on the principle of linear summation, which is not always satisfied by various physical and chemical factors. Algorithm BP is a completely different way to handle, without knowing or using any form of input / output relationship model, it is based on a certain learning rules to deal with the problem, especially for the causal relationship is not clear, the knowledge background is not clear, Uniqueness. We have used NN research quantitative structure-activity relationship, chemical pattern recognition, biochemical reaction modeling, complex mechanism analysis and multiple spectral analysis,