Python五行代码完成PCA

原理网上很多，不献丑。
代码包括生成有噪声数据和PCA降维。

效果：
PCA—3维降2维
PCA—二维转1维

代码：

import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D

def PCA(Data, k=2):
    A = Data-np.array([np.mean(Data,axis=1)]).T
    S = A @ A.T / (A.shape[1]-1)
    Sigma, U = np.linalg.eig(S)
    Dex = np.argsort(-Sigma)
    Sigma, U = Sigma[Dex], U[:,Dex]
    return U[:,:k],U[:,:k].T @ A

def MakeData2(s1):
    # x+y+z=1
    D = np.zeros((2, s1))
    k = np.random.randint(1,100)
    for i in range(s1):
        D[0,i] = np.random.randn()
        D[1,i] = k*D[0,i]+np.random.randn()*k
    return D

def MakeData3(s1):
    # x+y+z=1
    D = np.zeros((3, s1))
    for i in range(s1):
        D[0,i] = np.random.randn()
        D[1,i] = np.random.randn()
        D[2,i] = 1-D[0,i]-D[1,i]+np.random.randn()/2
    return D
def Plot2(A,Z):
    plt.figure()
    plt.scatter(A[0,:],A[1,:],c='b')
    plt.scatter(Z[0,:],Z[1,:],c='r')
    
def Plot3(A,Z):
    ax = plt.figure().add_subplot(111, projection = '3d')
    ax.scatter(A[0,:],A[1,:],A[2,:])
    ax.scatter(Z[0,:],Z[1,:],Z[2,:],c='r')

num = 100 # 样本数目
Data = MakeData3(num)
A = Data-np.array([np.mean(Data,axis=1)]).T
U, Y = PCA(A, 2)
Z = U @ Y
Plot3(A,Z)

Data = MakeData2(num)
A = Data-np.array([np.mean(Data,axis=1)]).T
U, Y = PCA(A, 1)
Z = U @ Y
Plot2(A,Z)