摘要
独立元分析是新引入化学计量学中的用于组分辨识的多元统计分析方法,奇异值分解是独立元分析的必要步骤。奇异值分解得到的基向量和独立元向量分别构成向量空间,当向量数目与实际体系数目一致时,两个向量空间是一致的,可以相互表述,利用幂等矩阵性质,可以显示出空间的差异。将奇异值分析结果与计算出的独立分量进行子空间差异对比,可以实现黑色体系组分数目的判断;经模拟和光谱分析数据证实,方法是准确、便捷的。
Independent Component Analysis (ICA) is a new multivariable statistical method which uses to recognize components in ch- emometrics, and the Singular Value Decomposition (SVD) is one initialization step in ICA. The base vectors of SVD and Independent Components of ICA respectively constitute two subspaces, the two subspaces are consistent and may be expressed mutually when the dimension number of subspaces equal to actual component number in system. Using idempotent nature, the difference of subspaces may be demonstrated, and comparing with the subspace of ICA and SVD, the component number of black system may be found out: Confirmed by the simulation dataset, and spectrum dataset, the method is accurate and convenient.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2007年第2期259-262,共4页
Computers and Applied Chemistry
基金
中国博士后科学基金(2005037583)
广西科学基金(桂科基0448010)
关键词
组分数目确定
独立元分析
子空间差异
component number determination, independent component analysis and subspace comparisons