CVCL/utils.py at master · cjhr95/CVCL · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
import numpy as np
import math
from sklearn.metrics import normalized_mutual_info_score, v_measure_score, adjusted_rand_score, accuracy_score
from sklearn import cluster
from sklearn.preprocessing import Normalizer
from scipy.optimize import linear_sum_assignment
from sklearn.decomposition import PCA


def normalize_multiview_data(data_views, row_normalized=True):
    '''The rows or columns of a matrix normalized '''
    norm2 = Normalizer(norm='l2')
    num_views = len(data_views)
    for idx in range(num_views):
        if row_normalized:
            data_views[idx] = norm2.fit_transform(data_views[idx])
        else:
            data_views[idx] = norm2.fit_transform(data_views[idx].T).T

    return data_views


def spectral_clustering(W, num_clusters):
    """
    Apply spectral clustering on W.
    # Arguments
    :param W: an affinity matrix
    :param num_clusters: the number of clusters
    :return: cluster labels.
    """
    # spectral = cluster.SpectralClustering(n_clusters=num_clusters, eigen_solver='arpack', affinity='precomputed',
    #                                       assign_labels='discretize')

    assign_labels='kmeans'
    spectral = cluster.SpectralClustering(n_clusters=num_clusters, eigen_solver='arpack', affinity='precomputed')
    spectral.fit(W)
    labels = spectral.fit_predict(W)

    return labels


def cal_spectral_embedding(W, num_clusters):

    D = np.diag(1 / np.sqrt(np.sum(W, axis=1) + math.e))
    # D1 = np.diag(np.power((np.sum(W, axis=1) + math.e), -0.5))
    Z = np.dot(np.dot(D, W), D)
    U, _, _ = np.linalg.svd(Z)
    eigenvectors = U[:, 0 : num_clusters]

    return eigenvectors


def cal_spectral_embedding_1(W, num_clusters):
    D = np.diag(np.power((np.sum(W, axis=1) + math.e), -0.5))
    L = np.eye(len(W)) - np.dot(np.dot(D, W), D)
    eigvals, eigvecs = np.linalg.eig(L)
    x_val = []
    x_vec = np.zeros((len(eigvecs[:, 0]), len(eigvecs[0])))
    for i in range(len(eigvecs[:, 0])):
        for j in range(len(eigvecs[0])):
            x_vec[i][j] = eigvecs[i][j].real
    for i in range(len(eigvals)):
        x_val.append(eigvals[i].real)
    # 选择前n个最小的特征向量
    indices = np.argsort(x_val)[: num_clusters]
    eigenvectors = x_vec[:, indices[: num_clusters]]

    return eigenvectors


def cal_l2_distances(data_view):
    '''
    calculate Euclidean distance
    tips: each row in data_view represents a sample
    '''
    num_samples = data_view.shape[0]
    dists = np.zeros((num_samples, num_samples))
    for i in range(num_samples):
        dists[i] = np.sqrt(np.sum(np.square(data_view - data_view[i]), axis=1)).T
    return dists


def cal_l2_distances_1(data_view):
    '''
    calculate Euclidean distance
    tips: each row in data_view represents a sample
    '''
    num_samples = data_view.shape[0]
    dists = np.zeros((num_samples, num_samples))
    for i in range(num_samples):
        for j in range(num_samples):
            dists[i][j] = np.sqrt(np.sum(np.square(data_view[i]-data_view[j])))

    return dists


def cal_squared_l2_distances(data_view):
    '''
    calculate squared Euclidean distance
    tips: each row in data_view represents a sample
    '''
    num_samples = data_view.shape[0]
    dists = np.zeros((num_samples, num_samples))
    for i in range(num_samples):
        dists[i] = np.sum(np.square(data_view - data_view[i]), axis=1).T
    return dists


def cal_squared_l2_distances_1(data_view):
    '''
    calculate squared Euclidean distance
    tips: each row in data_view represents a sample
    '''
    num_samples = data_view.shape[0]
    dists = np.zeros((num_samples, num_samples))
    for i in range(num_samples):
        for j in range(num_samples):
            dists[i][j] = np.sum(np.square(data_view[i]-data_view[j]))

    return dists


def cal_similiarity_matrix(data_view, k):
    '''
    calculate similiarity matrix
    '''
    num_samples = data_view.shape[0]
    dist = cal_squared_l2_distances(data_view)

    W = np.zeros((num_samples,num_samples), dtype=float)

    idx_set = dist.argsort()[::1]
    for i in range(num_samples):
        idx_sub_set = idx_set[i, 1:(k+2)]
        di = dist[i, idx_sub_set]
        W[i, idx_sub_set] = (di[k]-di) / (di[k] - np.mean(di[0:(k-1)]) + math.e)

    W = (W + W.T) / 2

    return W