SSE/main.py at main · SELGroup/SSE · 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
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
import sys

from SSE_hierarchical import PartitionTree_SSE


from queue import Queue

import scipy
from graph_construction import knn_affinity, knn_cosine_sim, generate_constraints_pairwise, generate_constraints_label, knn_k_estimating
from SSE_partitioning import FlatSSE
import numpy as np
from sklearn.metrics import adjusted_rand_score, normalized_mutual_info_score
from sklearn.preprocessing import StandardScaler, MinMaxScaler
# from utils import PartitionTree
from SSE_hierarchical import cal_dendrogram_purity
import argparse

parser = argparse.ArgumentParser()
parser.add_argument('--method', required=True, choices=['SSE_partitioning_pairwise', 'SSE_partitioning_label',
                                                         'SSE_partitioning_bio_pairwise', 'SSE_partitioning_bio_label', 'SSE_hierarchical'])
parser.add_argument('--dataset', required=True)
parser.add_argument('--constraint_ratio', type=float, required=True)
parser.add_argument('--constraint_weight', default=2.0, type=float)
parser.add_argument('--sigmasq', default=100, type=float, help='square of Gaussian kernel band width, i.e., sigma^2')
parser.add_argument('--exp_repeats', default=10, type=int)
parser.add_argument('--knn_constant', default=20, type=float)

# for hierarchical clustering
parser.add_argument('--hie_knn_k', default=5)
args = parser.parse_args()


def SSE_pairwise_clustering(path):
    data_f = scipy.io.loadmat(path)
    X = np.array(data_f['fea']).astype(float)
    X = MinMaxScaler().fit_transform(X)
    y = np.array(data_f['gnd']).astype(float).squeeze()
    n_instance = y.shape[0]
    n_cluster = np.unique(y).shape[0]
    knn_k = knn_k_estimating(n_cluster, n_instance, args.knn_constant)
    A, A_dense = knn_affinity(X, args.sigmasq, knn_k)
    ARIs = []
    NMIs = []
    for _ in range(args.exp_repeats):
        A_constraints = generate_constraints_pairwise(y, int(A.shape[0] * args.constraint_ratio), int(A.shape[0] * args.constraint_ratio), A_dense)
        flatSSE = FlatSSE(A, args.constraint_weight * A_constraints, len(np.unique(y)))
        y_pred_flatSSE = flatSSE.build_tree()
        ARI = adjusted_rand_score(y, y_pred_flatSSE)
        NMI = normalized_mutual_info_score(y, y_pred_flatSSE)
        ARIs.append(ARI)
        NMIs.append(NMI)
        print(path, ARI, NMI)
    print("average: {}\t{}\t{}\n".format(args.dataset, np.mean(ARIs), np.mean(NMIs)))

def SSE_label_clustering(path):
    data_f = scipy.io.loadmat(path)
    X = np.array(data_f['fea']).astype(float)
    X = MinMaxScaler().fit_transform(X)
    y = np.array(data_f['gnd']).astype(float).squeeze()
    n_instance = y.shape[0]
    n_cluster = np.unique(y).shape[0]
    knn_k = knn_k_estimating(n_cluster, n_instance, args.knn_constant)
    A, A_dense = knn_affinity(X, args.sigmasq, knn_k)
    ARIs = []
    NMIs = []
    for _ in range(args.exp_repeats):
        A_constraints = generate_constraints_label(y, int(A.shape[0] * args.constraint_ratio), int(A.shape[0] * args.constraint_ratio), A_dense)
        flatSSE = FlatSSE(A, args.constraint_weight * A_constraints, len(np.unique(y)))
        y_pred_flatSSE = flatSSE.build_tree()
        ARI = adjusted_rand_score(y, y_pred_flatSSE)
        NMI = normalized_mutual_info_score(y, y_pred_flatSSE)
        ARIs.append(ARI)
        NMIs.append(NMI)
        print(path, ARI, NMI)
    print("average: {}\t{}\t{}\n".format(args.dataset, np.mean(ARIs), np.mean(NMIs)))

def get_partition_from_2d(partition_tree, n_instances):
    root_id = partition_tree.root_id
    y_pred = np.zeros(n_instances)
    children = partition_tree.tree_node[root_id].children
    for index, child in enumerate(children):
        child = partition_tree.tree_node[child]
        partition = child.partition
        for vertex in partition:
            y_pred[vertex] = index
    return y_pred

def SSE_hierar_clustering(path, result_path=None):
    data_f = scipy.io.loadmat(path)
    X = np.array(data_f['fea']).astype(float)
    X = StandardScaler().fit_transform(X)
    y = np.array(data_f['gnd']).astype(float).squeeze()
    n_instance = y.shape[0]
    A, A_dense = knn_cosine_sim(X, args.hie_knn_k)
    DPs = []
    ARIs = []
    NMIs = []
    for _ in range(args.exp_repeats):
        A_constraints = generate_constraints_pairwise(y, int(A.shape[0] * args.constraint_ratio), int(A.shape[0] * args.constraint_ratio), A_dense)
        partitiontree_SSE = PartitionTree_SSE(A, args.constraint_weight*A_constraints)
        root_id, hierarchical_tree_node = partitiontree_SSE.build_coding_tree(2, mode='v1')
        y_pred = get_partition_from_2d(partitiontree_SSE, n_instance)
        DP = cal_dendrogram_purity(root_id, hierarchical_tree_node, n_instance, y)
        ARI = adjusted_rand_score(y, y_pred)
        NMI = normalized_mutual_info_score(y, y_pred)
        DPs.append(DP)
        ARIs.append(ARI)
        NMIs.append(NMI)
        print(path, DP, ARI, NMI)
    print(np.mean(DPs), np.mean(ARIs), np.mean(NMIs))
    if result_path is not None:
        with open(result_path, 'w') as f:
            f.write("DP:\t")
            for DP in DPs:
                f.write("{}\t".format(DP))
            f.write("average:\t{}\nARI:\t".format(np.mean(DPs)))
            for ARI in ARIs:
                f.write("{}\t".format(ARI))
            f.write("average:\t{}\nNMI:\t".format(np.mean(ARIs)))
            for NMI in NMIs:
                f.write("{}\t".format(NMI))
            f.write("average:\t{}\n".format(np.mean(NMIs)))

def SSE_pairwise_clustering_bio(path):
    data_f = scipy.io.loadmat(path)
    X = np.array(data_f['fea']).astype(float)
    X = MinMaxScaler().fit_transform(X)
    y = np.array(data_f['gnd']).astype(float).squeeze()
    n_instance = y.shape[0]
    n_cluster = np.unique(y).shape[0]
    knn_k = knn_k_estimating(n_cluster, n_instance, args.knn_constant)
    A, A_dense = knn_cosine_sim(X, knn_k)
    ARIs = []
    NMIs = []
    for _ in range(args.exp_repeats):
        A_constraints = generate_constraints_pairwise(y, int(A.shape[0] * args.constraint_ratio),
                                                      int(A.shape[0] * args.constraint_ratio), A_dense)
        flatSSE = FlatSSE(A, args.constraint_weight * A_constraints, len(np.unique(y)), mustlink_first=True)
        y_pred_flatSSE = flatSSE.build_tree()
        ARI = adjusted_rand_score(y, y_pred_flatSSE)
        NMI = normalized_mutual_info_score(y, y_pred_flatSSE)
        ARIs.append(ARI)
        NMIs.append(NMI)
        print(path, ARI, NMI)
    print(np.mean(ARIs), np.mean(NMIs))

def SSE_label_clustering_bio(path):
    data_f = scipy.io.loadmat(path)
    X = np.array(data_f['fea']).astype(float)
    X = MinMaxScaler().fit_transform(X)
    y = np.array(data_f['gnd']).astype(float).squeeze()
    n_instance = y.shape[0]
    n_cluster = np.unique(y).shape[0]
    knn_k = knn_k_estimating(n_cluster, n_instance, args.knn_constant)
    A, A_dense = knn_cosine_sim(X, knn_k)
    ARIs = []
    NMIs = []
    for _ in range(args.exp_repeats):
        A_constraints = generate_constraints_label(y, int(A.shape[0] * args.constraint_ratio),
                                                      int(A.shape[0] * args.constraint_ratio), A_dense)
        flatSSE = FlatSSE(A, args.constraint_weight * A_constraints, len(np.unique(y)), mustlink_first=True)
        y_pred_flatSSE = flatSSE.build_tree()
        ARI = adjusted_rand_score(y, y_pred_flatSSE)
        NMI = normalized_mutual_info_score(y, y_pred_flatSSE)
        ARIs.append(ARI)
        NMIs.append(NMI)
        print(path, ARI, NMI)
    print(np.mean(ARIs), np.mean(NMIs))

if __name__=='__main__':
    if args.method == "SSE_partitioning_pairwise":
        path = "./datasets/clustering/{}.mat".format(args.dataset)
        SSE_pairwise_clustering(path)
    elif args.method == "SSE_partitioning_label":
        path = "./datasets/clustering/{}.mat".format(args.dataset)
        SSE_label_clustering(path)
    elif args.method == "SSE_partitioning_bio_pairwise":
        path = "./datasets/RNA-seq/{}.mat".format(args.dataset)
        SSE_pairwise_clustering_bio(path)
    elif args.method == "SSE_partitioning_bio_label":
        path = "./datasets/RNA-seq/{}.mat".format(args.dataset)
        SSE_label_clustering_bio(path)
    elif args.method == "SSE_hierarchical":
        path = "./datasets/hierarchical/{}.mat".format(args.dataset)
        SSE_hierar_clustering(path)