调包计算拉普拉斯矩阵

import numpy as np
from sklearn import datasets
from sklearn.neighbors import NearestNeighbors, kneighbors_graph

X, y = datasets.load_iris(return_X_y=True)

# -------------第一种方法----------------------
A = kneighbors_graph(X, n_neighbors=5, mode="connectivity",include_self=False)
R = A.toarray()
S1 = R + R.T
S1[S1 >= 1] = 1 # 近邻权重矩阵
D = np.diag(np.sum(S1, axis=1))  # the diagonal matrix
L = D - S1 # the Laplacian matrix


# ---------------------------------------------
neigh = NearestNeighbors(n_neighbors=5)
neigh.fit(X=X)
nei = neigh.kneighbors()[1]  # the neighbors of each instance

S2 = np.zeros((len(y), len(y)))
for i in range(len(y)):
    for j in range(len(y)):
        if i in nei[j] or j in nei[i]:
            S2[i,j] = 1
D = np.diag(np.sum(S2, axis=1))  # the diagonal matrix
L = D - S2 # the Laplacian matrix

# ==========验证两种方法得到的近邻权重矩阵是相同的==============
print("验证1：",np.sum(S1 - S2))