Stanford ml Open Course Notes 14-Principal Component Analysis

In the previous note, we introduced the factor analysis model. The factor analysis model uses the Hidden variable Z in the D-dimension sub-space to fit the training data. In fact, the factor analysis model is a method for dimensionality reduction of data, it uses the EM algorithm to estimate parameters based on a probability model.

This article mainly introduces PCA (Principal Component Analysis), which is also a dimensionality reduction method. However, this method is more direct and can be used to perform dimensionality reduction simply by calculating feature vectors. The video corresponding to this article is the 14th videos of the open class. The first half of this video is the EM solution of the factor analysis model, which has been written into note 13. This article is only the second half of the note, therefore, the content is small.