-
Notifications
You must be signed in to change notification settings - Fork 0
/
clustering_strategies.py
44 lines (34 loc) · 1.44 KB
/
clustering_strategies.py
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
import operator
from sklearn import cluster
import utils
import numpy as np
def strategy_template(adj_list, n, method):
G_degrees = dict(map(lambda (k, v): (k, len(v)), adj_list.iteritems()))
i_matrix = utils.adjacency_list_to_incidence_matrix(adj_list)
clusters = method.fit_predict(i_matrix)
degree_sorted_clusters = []
for i in range(8):
curr_cluster = np.where(clusters==i)[0]
curr_cluster_degrees = dict((k, G_degrees[k]) for k in curr_cluster)
degree_sorted_cluster = [node for (node, degree) in sorted(curr_cluster_degrees.items(), key=lambda x: x[1], reverse=True)]
degree_sorted_clusters.append(degree_sorted_cluster)
size_degree_sorted_clusters = sorted(degree_sorted_clusters, key=lambda x: len(x), reverse=True)
seed_nodes = []
n_left = n
for cluster in size_degree_sorted_clusters:
if n_left == 0:
break
num = min(n_left, len(cluster))
seed_nodes += cluster[0:num]
n_left -= num
return seed_nodes
def spectral_clustering(adj_list, n):
"""
Returns highest degree nodes from the largest clusters defined by spectral clustering
"""
return strategy_template(adj_list, n, cluster.SpectralClustering(affinity='precomputed'))
def k_means_clustering(adj_list, n):
"""
Returns highest degree nodes from the largest cluster defined by k-means clustering
"""
return strategy_template(adj_list, n, cluster.KMeans())