python 矩阵特征值分解_讲一下numpy的矩阵特征值分解与奇异值分解

1、特征值分解

主要还是调包：

from numpy.linalg import eig

特征值分解：  A = P*B*PT 当然也可以写成 A = QT*B*Q 其中B为对角元为A的特征值的对角矩阵，P=QT，

首先A得对称正定，然后才能在实数域上分解，

>>> A = np.random.randint(-10,10,(4,4))>>>A

array([[6, 9, -10, -1],

[5, 9, 5, -5],

[-8, 7, -4, 4],

[-1, -9, 0, 6]])>>> C =np.dot(A.T, A)>>>C

array([[126, 52, -3, -69],

[52, 292, -73, -80],

[-3, -73, 141, -31],

[-69, -80, -31, 78]])>>> vals, vecs =eig(C)>>>vals

array([357.33597086, 174.10172008, 8.84429957, 96.71800949])>>>vecs

array([[-0.28738314, -0.51589436, -0.38221983, -0.71075449],

[-0.87487263, 0.12873861, -0.24968051, 0.39456798],

[0.2572149 , -0.69304313, -0.33950158, 0.58161018],

[0.29300052, 0.48679627, -0.82237845, -0.02955945]