import numpy as np
import sklearn.metrics as metrics
import scipy.io
from scipy.spatial.distance import pdist, squareform
from scipy.cluster import hierarchy
import matplotlib.pyplot as plt
plt.xkcd()
import networkx as nx
%matplotlib inline

G = nx.karate_club_graph()
cliqComm = nx.k_clique_communities(G, 3)
cliqComm=list(cliqComm)
cliqComm

colorList=np.ones([G.order(),])
for c in xrange(len(cliqComm)):
    for i in cliqComm[c]:
        colorList[i] += c+1
print colorList    

nx.draw(G, 
        cmap = plt.get_cmap('Reds'), 
        node_color=colorList, 
        node_size=500, 
        alpha=0.8, 
        with_labels=True)

data = scipy.io.loadmat('network3.mat')
A = data['A'].astype('float')
plt.spy(A)
comm = data['Comm']

G = nx.Graph(A)
# pos = nx.spring_layout(G, scale = 5, iterations=100)

nx.draw(G, 
        pos, 
        node_color=comm,
        alpha = 0.6,
        with_labels = False)

D = pdist(A, metric = 'jaccard')
hc = hierarchy.average(D)
Z = hierarchy.dendrogram(hc)

comm2 = hierarchy.fcluster(hc, 4, criterion='maxclust')
comm = comm.flatten()
metrics.adjusted_rand_score(comm, comm2)

nx.draw(G, 
        pos, 
        node_color=comm2,
        alpha = 0.6,
        with_labels = False)