import networkx as nx
import matplotlib.pyplot as plt
plt.xkcd()
import numpy as np
%matplotlib inline

G = nx.read_gml(path = 'karate.gml')
nx.draw_networkx(G)

k = nx.degree(G)
plt.figure(1, figsize=(10,7))
coord = nx.spring_layout(G)
nx.draw_networkx(G,
                 pos=coord,
                 nodelist=k.keys(),
                 node_size = [d*100 for d in k.values()],
                 node_color=k.values(),
                 font_size=8,
                 cmap=plt.cm.Reds,
                 )

# load adjacency matrix
A = nx.to_numpy_matrix(G, dtype=np.bool) 
k = G.degree()
# alternatevly you can find k as k = np.sum(A, axis=0)

# sort nodes according to degree
id_sort = np.array(sorted(k, key = k.get, reverse=True))
k_sort = np.array(k.values())
k_sort = k_sort[id_sort-1]

# show adjacency matrix
plt.figure(1, figsize=(6, 6))
plt.imshow(A,
           cmap="Greys",
           interpolation="none"
           )
plt.title('Zachary\'s club adjacency matrix', fontsize=14)

# show node degrees
plt.figure(1, figsize=(10, 5))
plt.bar(np.arange(34), k_sort)
plt.xticks(np.arange(34), id_sort)
plt.margins(0.2)
plt.ylabel('Degree')
plt.xlabel('Node id')

# Write your code here
#
#
#

# load data
D = np.loadtxt('bikinis.txt')
y = D[:,0]
X = D[:,1:]

# Find regression coefs
Beta = np.dot(np.dot(np.linalg.inv(np.dot(X.T, X)), X.T), y)

# Make predictions
yt = X.dot(Beta)

# Calculate residuals
res = y - yt

# Plot <resuduals against preditions>
plt.figure(figsize=(10,10))
plt.plot(yt, res, '+b')

# Choose 2 dims and plot predictions agains the line
X1 = X[:,0]
idx = np.argsort(X1)
yt1 = X1[idx].dot(Beta[0])

fig, ax = plt.subplots()
fig.set_size_inches(7,7)
ax.plot(X1, y, '.r', label='Data') 
ax.plot(X1[idx], yt1, 'b', label='Line')
plt.xlabel('$x_1$')
plt.ylabel('$y$')
ax.legend(loc='upper left', shadow=True)

# Write your code here
#
#
#

# Loading network
G = nx.read_gml('network.gml')

# Draw it!
pos = nx.spring_layout(G, dim = 2, iterations=100)
plt.figure(figsize =(7,7))
nx.draw(G, pos)

print 'Network order {0:d}, Network size {1:d}'.format(G.order(), G.size())

# Degree distr
k = G.degree()
k = np.array(k.values())
n, bins, patches = plt.hist(k, bins = 10, facecolor='g')

# Find nodes with highest degree
idx = np.argsort(k)
idx = idx[::-1]
k_sorted = k[idx]
v1 = G.node[idx[0]]
print v1['label']

# Find connected components
CC = nx.connected_components(G)
for v in CC[0]:
    print G.node[v]

SG = G.subgraph(CC[0])
nx.radius(SG)