摘要
混合因子分析是一种对具有复杂结构的多维数据建立模型的方法.本文提出了一种进行混合因子分析的重新抽样方法.当给定一组数据样本时,我们首先建立样本概率分布的混合高斯模型,然后为每一个高斯混合项重新抽取新的数据样本,在新的样本上再对每一个高斯混合项进行因子分析.与已有的算法相比较,避免了计算各个高斯混合项在每个样本值之下的后验概率,又减少了进行因子分析时参与计算的数据样本的数量.
The mixtures of factor analyzers are able to model complex data structures through a combination of the factor analysis model and the Gaussian mixture model. In this paper, a resampling method for the mixtures of factor analyzers is proposed. After approximating the probability distribution density of the data by the Gaussian mixture model, we draw new samples for each of the component Gaussians with its own parameters separately, then on the new samples the factor analysis is performed for each component Gaussians. We also implement this method with the EM algorithm and the good performance of the method is illustrated by an example.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2002年第12期1873-1875,共3页
Acta Electronica Sinica
基金
国家自然科学基金(No.60073053)