协方差矩阵的计算方法

来源：互联网发布：阿里云企业邮箱免费吗编辑：程序博客网时间：2024/04/30 16:57

C.M Bishop的《Pattern Recognition and Machine Learning》12章介绍PCA，式（12.3）计算样本的协方差矩阵，写了个简单的python代码计算下，并与numpy中的cov函数对比下，结果一致

python代码如下：

import numpy as npdef cov(data):    mean_ = np.mean(data, axis = 0)    data = data - mean_    cov_mat = data.T.dot(data) / (data.shape[0] - 1)    return cov_matif __name__ == '__main__':    a = np.random.random((10,2))        cov_mat = np.cov(a, rowvar = 0)    print 'covariance matrix using Numpy'    print cov_mat    my_cov_mar = cov(a)    print 'covariance matrix from Bishop book (eq12.3).<<Pattern Recognition and Machine Learning>>'    print my_cov_mar

计算结果为：
covariance matrix using Numpy
[[ 0.1063101 -0.00794282]
[-0.00794282 0.07562934]]
covariance matrix from Bishop book (eq12.3).<<Pattern Recognition and Machine Learning>>
[[ 0.1063101 -0.00794282]
[-0.00794282 0.07562934]]