CV-EigenModel&PCA

Variance(1D data)：A single number (scalar) encoding the spread about the mean μ (also the square of standard deviation).

方差用来描述一维数据，描述了数据与其平均值的偏离程度，是标准差的平方。

Covariance Matrix(any-D data): A dxd symmetric matrix encoding spread about the mean μ, and how each dimension varies with respect to other dimensions. Written C.

方差是用来描述一维数据的，针对多维数据，使用协方差矩阵，能够展现各个维度数据之间的关系。

 

Decomposition of Covariance

It is possible to decompose the covariance matrix C. (using the Eigenvalue Decomposition(EVD))

C = U V U'   

U = Matrix of Eigenvecotrs    V = Matrix of Eigenvalues

Each eigenvector has a corresponding eigenvalue. The eigenvector corresponding to the largest eigenvalue is called the principal eigenvector. It points in the direction of greatest variance. Eigenvectors form a data-dependent  basis set for a reference f