def plot_graph(graph, adj):
""" Use the package networkx to produce a diagrammatic plot of the graph, with
the nodes in the graph colored according to their current labels.
Note that only 20 unique colors are available for the current color map,
so common colors across nodes may be coincidental.
Parameters
----------
graph : Tuple[Node, ...]
The graph to plot
adj : numpy.ndarray, shape=(N, N)
The adjacency-matrix for the graph. Nonzero entries indicate
the presence of edges.
Returns
-------
Tuple[matplotlib.fig.Fig, matplotlib.axis.Axes]
The figure and axes for the plot."""
import networkx as nx
import numpy as np
import matplotlib.cm as cm
import matplotlib.pyplot as plt
g = nx.Graph()
for n, node in enumerate(graph):
g.add_node(n)
# construct a network-x graph from the adjacency matrix: a non-zero entry at adj[i, j]
# indicates that an egde is present between Node-i and Node-j. Because the edges are
# undirected, the adjacency matrix must be symmetric, thus we only look ate the triangular
# upper-half of the entries to avoid adding redundant nodes/edges
g.add_edges_from(zip(*np.where(np.triu(adj) > 0)))
# we want to visualize our graph of nodes and edges; to give the graph a spatial representation,
# we treat each node as a point in 2D space, and edges like compressed springs. We simulate
# all of these springs decompressing (relaxing) to naturally space out the nodes of the graph
# this will hopefully give us a sensible (x, y) for each node, so that our graph is given
# a reasonable visual depiction
pos = nx.spring_layout(g)
# make a mapping that maps: node-lab -> color, for each unique label in the graph
color = list(
iter(cm.Vega20b(np.linspace(0, 1, len(set(i.label for i in graph))))))
color_map = dict(zip(sorted(set(i.label for i in graph)), color))
colors = [
color_map[i.label] for i in graph
] # the color for each node in the graph, according to the node's label
# render the visualization of the graph, with the nodes colored based on their labels!
fig, ax = plt.subplots()
nx.draw_networkx_nodes(g,
pos=pos,
ax=ax,
nodelist=range(len(graph)),
node_color=colors)
nx.draw_networkx_edges(g, pos, ax=ax, edgelist=g.edges())
return fig, ax
def plot_graph(graph, adj, labels):
""" Use the package networkx to produce a diagrammatic plot of the graph, with
the nodes in the graph colored according to their current labels.
Note that only 20 unique colors are available for the current color map,
so common colors across nodes may be coincidental.
Parameters
----------
graph : Tuple[Node, ...]
The graph to plot
adj : numpy.ndarray, shape=(N, N)
The adjacency-matrix for the graph. Nonzero entries indicate
the presence of edges.
Returns
-------
Tuple[matplotlib.fig.Fig, matplotlib.axis.Axes]
The figure and axes for the plot."""
g = nx.Graph()
for n, node in enumerate(graph):
g.add_node(n)
g.add_edges_from(zip(*np.where(np.triu(adj) > 0)))
pos = nx.spring_layout(g)
color = list(
iter(cm.Vega20b(np.linspace(0, 1, len(set(i.label for i in graph))))))
color_map = dict(zip(sorted(set(i.label for i in graph)), color))
colors = [color_map[i.label] for i in graph]
fig, ax = plt.subplots()
nx.draw_networkx_nodes(g,
pos=pos,
ax=ax,
nodelist=range(len(graph)),
node_color=colors)
nx.draw_networkx_edges(g, pos, ax=ax, edgelist=g.edges())
nx.draw_networkx_labels(g, pos=pos, ax=ax, labels=labels)
return fig, ax
import numpy as np
import scipy.cluster.hierarchy as sch
import matplotlib
matplotlib.use('Agg')
import matplotlib.pylab as plt
from matplotlib import cm
import fastcluster as fcl
import glob
import os
import cPickle
import PIL
(npArray, D, Z1, names) = cPickle.load(open("clusterstate.pickle"))
im = PIL.Image.fromarray(cm.Vega20b(D, bytes=True))
im.save('heatmap.png')
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
import numpy as np
from matplotlib import cm
with open('data-user-groups.csv', 'r') as in_file:
data = []
line = in_file.readline()
for d in line.strip().split(','):
data.append(int(d))
fig = plt.figure(figsize=(8, 4))
ax = fig.add_subplot(111, aspect='equal')
labels = ('Watch', 'Dismiss only', 'No activity')
cs = cm.Vega20b(np.arange(3) / 3.)
patches, texts, autotexts = plt.pie(data,
labels=labels,
autopct='%1.1f%%',
colors=cs,
startangle=90)
autotexts[0].set_color('w')
plt.tight_layout()
with PdfPages('user-groups-pie.pdf') as pdf:
pdf.savefig(fig)
plt.close()