In [1]:
import os
os.chdir('/Users/Tony/Documents/Git Folder/seelviz/Jupyter/DownsampleGraphML')
from argparse import ArgumentParser
from collections import OrderedDict
from subprocess import Popen
from scipy.stats import gaussian_kde
from matplotlib.backends.backend_pdf import PdfPages
import numpy as np
import nibabel as nb
import networkx as nx
import os
import pickle
import matplotlib.pyplot as plt
In [2]:
# Change the filename below to run different graphml files
filename = 'Fear199localeq.5000.graphml'
G = nx.read_graphml(filename)
In [3]:
def scan_statistic(G, i):
"""
Computes scan statistic-i on a set of graphs
Required Parameters:
G: graph
i:
- which scan statistic to compute
"""
tmp = np.array(())
for n in G.nodes():
sg = nx.ego_graph(G, n, radius=i)
tmp = np.append(tmp, np.sum([sg.get_edge_data(e[0], e[1])['weight']
for e in sg.edges()]))
ss = tmp
return ss
In [4]:
def density(data):
"""
Computes density for metrics which return vectors
Required parameters:
data:
- Dictionary of the vectors of data
"""
dens = gaussian_kde(data)
xs = np.linspace(0, 1.2*np.max(data), 1000)
density = dens.pdf(xs)
return {"xs": xs, "pdfs": density}
In [5]:
def write(outdir, metric, data, atlas):
"""
Write computed derivative to disk in a pickle file
Required parameters:
outdir:
- Path to derivative save location
metric:
- The value that was calculated
data:
- The results of this calculation
atlas:
- Name of atlas of interest as it appears in the directory titles
"""
of = open(outdir + '/' + atlas + '_' + metric + '.pkl', 'wb')
pickle.dump({metric: data}, of)
of.close()
In [6]:
# NNZ - Number of non-zero edges
print len(nx.edges(G))
write('../pickels', 'non-zero_edges', nx.edges(G), filename)
In [8]:
# Degree sequence
temp_deg = np.array(nx.degree(G).values())
deg = density(temp_deg)
# print deg
xs = deg['xs']
pdfs = deg['pdfs']
# Autoscale Output Image
pdfs = [i for i in pdfs if i >= 0.0001]
xs = xs[0: len(pdfs)]
xsMAX = np.amax(xs)
pdfsMAX = np.amax(pdfs)
# Name of Output File
pp = PdfPages('DegreeSequencefor' + filename + '.pdf')
# Plot File
plt.axis([0, xsMAX + 0.001, 0, pdfsMAX + 0.1])
plt.xlabel('Degree sequence for ' + filename)
plt.ylabel('PDFs')
plt.title(r'Degree sequence vs PDFs for ' + filename)
fig1 = plt.plot(xs, pdfs)
# Tweak spacing to prevent clipping of ylabel
plt.subplots_adjust(left=0.15)
pp.savefig()
pp.close()
plt.clf()
# Make pickel
write('../pickels', 'degree sequence', deg, filename)
In [9]:
# Compute Edge Weight
temp_ew = (G.get_edge_data(e[0], e[1])['weight'] for e in G.edges())
## ew = density(temp_ew) "ValueError: `dataset` input should have multiple elements."
print temp_ew
# We don't have edge weights so this makes sense
# Make pickel
# write('../pickels', 'edge weight', temp_ew, filename)
In [10]:
# Clustering Coefficients
nxc = nx.clustering
temp_cc = nxc(G).values()
ccoefs = density(temp_cc)
xscc = ccoefs['xs']
pdfscc = ccoefs['pdfs']
# Autoscale Output Image
pdfscc = [i for i in pdfscc if i >= 0.0001]
xscc = xscc[0: len(pdfscc)]
xsccMAX = np.amax(xscc)
pdfsccMAX = np.amax(pdfscc)
# Name of Output File
pp2 = PdfPages('ClusteringCoefficientsfor' + filename + '.pdf')
# Plot File
plt.axis([0, xsccMAX + 0.01, 0, (pdfsccMAX + 0.1)])
plt.xlabel('Clustering values for ' + filename)
plt.ylabel('PDFs')
plt.title(r'Clustering values vs PDFs for ' + filename)
fig2 = plt.plot(xscc, pdfscc)
# Tweak spacing to prevent clipping of ylabel
plt.subplots_adjust(left=0.15)
pp2.savefig()
pp2.close()
plt.clf()
# Make pickel
write('../pickels', 'clustering coefficient', ccoefs, filename)
In [11]:
# Scan Statistic-1
temp_ss1 = scan_statistic(G, 1)
ss1 = density(temp_ss1)
xsss = ss1['xs']
pdfsss = ss1['pdfs']
# Autoscale Output Image
pdfsss = [i for i in pdfsss if i >= 0.0001]
xsss = xsss[0: len(pdfsss)]
xsssMAX = np.amax(xsss)
pdfsssMAX = np.amax(pdfsss)
pp3 = PdfPages('ScanStatistic1for' + filename + '.pdf')
plt.axis([0, xsssMAX + 1, 0, pdfsssMAX + 0.01])
plt.xlabel('Scan Statistic values for' + filename)
plt.ylabel('PDFs')
plt.title(r'Scan Statistic values vs PDFs for ' + filename)
fig3 = plt.plot(xsss, pdfsss)
# Tweak spacing to prevent clipping of ylabel
plt.subplots_adjust(left=0.15)
pp3.savefig()
pp3.close()
plt.clf()
# Make pickel
write('../pickels', 'scan statistic-1', ss1, filename)
In [17]:
# EigenValues
laplacian = nx.normalized_laplacian_matrix(G)
eigs = np.sort(np.linalg.eigvals(laplacian.A))[::-1]
pp = PdfPages('Eigenvaluesfor ' + filename + '.pdf')
num_bins = 50
# the histogram of the data
n, bins, patches = plt.hist(eigs, num_bins, facecolor='green', alpha=0.5)
plt.axis([0, 3, 0, 500])
plt.xlabel('Eigenvalues for ' + filename)
plt.ylabel('PDFs')
plt.title(r'Histogram of Eigenvalues for ' + filename)
# Tweak spacing to prevent clipping of ylabel
plt.subplots_adjust(left=0.15)
pp.savefig()
pp.close()
In [13]:
#Betweenness Centrality
nxbc = nx.algorithms.betweenness_centrality
temp_bc = (nxbc(G).values())
centrality = density(temp_bc)
xsbetween = centrality['xs']
pdfsbetween = centrality['pdfs']
# Autoscale Output Image
pdfsbetween = [i for i in pdfsbetween if i >= 0.0001]
xsbetween = xsbetween[0: len(pdfsbetween)]
xsbetweenMAX = np.amax(xsbetween)
pdfsbetweenMAX = np.amax(pdfsbetween)
pp4 = PdfPages('Betweenness Centrality' + filename + '.pdf')
# Plot
plt.axis([0, xsbetweenMAX, 0, pdfsbetweenMAX])
plt.xlabel('Betweenness Centrality values for ' + filename)
plt.ylabel('PDFs')
plt.title(r'Betweenness Centrality values vs PDFs for ' + filename)
plt.plot(xsbetween, pdfsbetween)
# Tweak spacing to prevent clipping of ylabel
plt.subplots_adjust(left=0.15)
pp4.savefig()
pp4.close()
plt.clf()
# Make pickel
write('../pickels', 'betweenness centrality', centrality, filename)
In [ ]: