def calculate(index):
"""
takes a pickled graph. If it's been analyzed,
returns the result of the analysis. Otherwise,
it analyzes the graph for clustering, knots,
number of vertices and number of edges
"""
filename = '../Graphs/' + index[6:] + '_graph.txt'
try:
results = load_object_from_file('../Results/' + index[6:] + '_results.txt')
print "Loaded %s" %(index[6:])
except IOError:
g = load_object_from_file(filename)
results = analyze_graph(g,index[6:])
save_object_to_file(results, '../Results/' + index[6:] + '_results.txt')
print "Saved %s" %(index[6:])
except:
print 'Something bad happened....'
return results
def GetResultList():
"""
Returns a list of tuples containing the list of:
clustering coefficients
has_knot
number of vertices
number of edges
"""
indices = load_object_from_file('../Graphs/indices.txt')
cs = []
k = []
vs = []
es = []
for index in indices:
try:
result = load_object_from_file('../Results/' + index[6:] + '_results.txt')
cs.append(result.clustering)
k.append(result.has_knot)
vs.append(result.vertices)
es.append(result.edges)
except:
pass
return cs, k, vs, es
def MultiThreadit():
resultDict = dict() #resulting super awesome dictionary
def cb(IndexResults):
"""
Puts the analysis result in the return dictionary
"""
resultDict[IndexResults.name] = IndexResults
indices = load_object_from_file('../Graphs/indices.txt')
po = Pool() #pool of processes
for index in indices: #each process takes a graph, asynchronously
po.apply_async(calculate,(index,),callback=cb)
po.close()
po.join()
print resultDict
return resultDict
def three_graph_subplot():
indices = load_object_from_file('../Graphs/indices.txt')
graphs = []
for index in indices:
try:
results = load_object_from_file('../Results/' + index[6:] + '_results.txt')
if results.vertices >= 5200:
graph = load_object_from_file('../Graphs/' + index[6:] + '_graph.txt')
graphs.append((index[6:], graph))
except:
pass
subplot_count = 1
name, graph = graphs[0]
#build list of in degree, out degree, total of each vertex
ins, outs, totals = [], [], []
for vertex in graph.vertices():
deg_in = graph.in_degree(vertex)
deg_out = graph.out_degree(vertex)
deg_tot = deg_in + deg_out
ins.append(deg_in)
outs.append(deg_out)
totals.append(deg_tot)
in_out = []
for item in zip(ins,outs):
if item[0] < 1 or item[1] < 1:
continue
else:
in_out.append(item[0]/float(item[1]))
c = Cdf.MakeCdfFromList(in_out)
x, y = c.Render()
pyplot.plot(x,y,'o')
pyplot.xscale('log')
print numpy.mean(in_out)
print numpy.median(in_out)
#~ pyplot.yscale('log')
#~ pyplot.plot(ins,outs,'o')
#~ pyplot.xlabel('in-degree')
#~ pyplot.ylabel('out-degree')
#~ pyplot.xscale('log')
#~ pyplot.yscale('log')
#~ print correlation.Corr(ins,outs)
#~ print correlation.SpearmanCorr(ins,outs)
#~ xs_log0 = []
#~ ys_log0 = []
#~ for x in ins:
#~ if x <= 0:
#~ xs_log0.append(.00001)
#~ else:
#~ xs_log0.append(math.log(x))
#~ for y in outs:
#~ if y <= 0:
#~ ys_log0.append(.00001)
#~ else:
#~ ys_log0.append(math.log(y))
#~ print correlation.Corr(xs_log0,ys_log0)
#~ print correlation.SpearmanCorr(xs_log0,ys_log0)
#~ coefs = numpy.lib.polyfit(xs_log0, ys_log0, 1)
#~ print coefs
#~ fit_y0 = numpy.lib.polyval(coefs, xs_log0)
#~ fit_y_log0 = [math.exp(1) ** f for f in fit_y0]
#~ pyplot.plot(ins, fit_y_log0,'r--',linewidth=4)
#create a pmf of degrees
#~ pmf0 = Pmf.MakePmfFromList(ins)
#~ xs0, ys0 = pmf0.Render()
#~
#~ pmf1 = Pmf.MakePmfFromList(outs)
#~ xs1, ys1 = pmf1.Render()
#~
#~ pmf2 = Pmf.MakePmfFromList(totals)
#~ xs2, ys2 = pmf2.Render()
#~
#~ #convert to log, so we can find line of best fit
#~ xs_log0 = []
#~ ys_log0 = []
#~ for x in xs0:
#~ if x <= 0:
#~ xs_log0.append(.00001)
#~ else:
#~ xs_log0.append(math.log(x))
#~ for y in ys0:
#~ if y <= 0:
#~ ys_log0.append(.00001)
#~ else:
#~ ys_log0.append(math.log(y))
#~ coefs = numpy.lib.polyfit(xs_log0, ys_log0, 1)
#~ fit_y0 = numpy.lib.polyval(coefs, xs_log0)
#~
#~ #convert to log, so we can find line of best fit
#~ xs_log1 = []
#~ ys_log1 = []
#~ for x in xs1:
#~ if x <= 0:
#~ xs_log1.append(.00001)
#~ else:
#~ xs_log1.append(math.log(x))
#~ for y in ys1:
#~ if y <= 0:
#~ ys_log1.append(.00001)
#~ else:
#~ ys_log1.append(math.log(y))
#~ coefs = numpy.lib.polyfit(xs_log1, ys_log1, 1)
#~ fit_y1 = numpy.lib.polyval(coefs, xs_log1)
#~
#~ #convert to log, so we can find line of best fit
#~ xs_log2 = []
#~ ys_log2 = []
#~ for x in xs2:
#~ if x <= 0:
#~ xs_log2.append(.00001)
#~ else:
#~ xs_log2.append(math.log(x))
#~ for y in ys2:
#~ if y <= 0:
#~ ys_log2.append(.00001)
#~ else:
#~ ys_log2.append(math.log(y))
#~ coefs = numpy.lib.polyfit(xs_log2, ys_log2, 1)
#~ fit_y2 = numpy.lib.polyval(coefs, xs_log2)
#transform fit line, to plot it on log-log scale
#~ pyplot.subplot(1,3,1)
#~ fit_y_log0 = [math.exp(1) ** f for f in fit_y0]
#~ pyplot.plot(xs0, ys0, 'o')
#~ pyplot.plot(xs0, fit_y_log0,'r--',linewidth=4)
#~ pyplot.xscale('log')
#~ pyplot.yscale('log')
#~ pyplot.ylabel('P(k)',fontsize=25)
#~
#~ pyplot.subplot(1,3,2)
#~ fit_y_log1 = [math.exp(1) ** f for f in fit_y1]
#~ pyplot.plot(xs1, ys1, 'o')
#~ pyplot.plot(xs1, fit_y_log1,'r--',linewidth=4)
#~ pyplot.xscale('log')
#~ pyplot.yscale('log')
#~ pyplot.xlabel('k',fontsize=25)
#~
#~ pyplot.subplot(1,3,3)
#~ fit_y_log2 = [math.exp(1) ** f for f in fit_y2]
#~ pyplot.plot(xs2, ys2, 'o')
#~ pyplot.plot(xs2, fit_y_log2,'r--',linewidth=4)
#~ pyplot.xscale('log')
#~ pyplot.yscale('log')
#~ if subplot_count == 1:
#~ pyplot.ylabel('P(k)',fontsize=25)
#~ if subplot_count == 2:
#~ pyplot.xlabel('k',fontsize=25)
#pyplot.show()
#~ title = 'total_degree_' + name
#~ subplot_count += 1
pyplot.show()
def compare_wikipedia_to_ba():
"""
For wikipedia articles of at least 4000 vertices, builds a graph of
k vs P(k), and finds line of best fit. Also builds this same graph
for a Barabasi Albert graph on the same number of vertices.
"""
indices = load_object_from_file('../Graphs/indices.txt')
for index in indices:
try:
results = load_object_from_file('../Results/' + index[6:] + '_results.txt')
if results.vertices >= 3700:
#~ print index[6:]
graph = load_object_from_file('../Graphs/' + index[6:] + '_graph.txt')
graphs.append((index[6:], graph))
except:
pass
for name, graph in graphs[:3]:
#build list of in degree, out degree, total of each vertex
ins, outs, totals = [], [], []
for vertex in graph.vertices():
deg_in = graph.in_degree(vertex)
deg_out = graph.out_degree(vertex)
deg_tot = deg_in + deg_out
ins.append(deg_in)
outs.append(deg_out)
totals.append(deg_tot)
#create a pmf of degrees
pmf = Pmf.MakePmfFromList(outs)
xs, ys = pmf.Render()
#convert to log, so we can find line of best fit
xs_log = []
ys_log = []
for x in xs:
if x <= 0:
xs_log.append(.00001)
else:
xs_log.append(math.log(x))
for y in ys:
if y <= 0:
ys_log.append(.00001)
else:
ys_log.append(math.log(y))
coefs = numpy.lib.polyfit(xs_log, ys_log, 1)
fit_y = numpy.lib.polyval(coefs, xs_log)
print coefs
#transform fit line, to plot it on log-log scale
fit_y_log = [math.exp(1) ** f for f in fit_y]
pyplot.clf()
pyplot.plot(xs, ys, 'o')
pyplot.plot(xs, fit_y_log,'r--',linewidth=4)
pyplot.xscale('log')
pyplot.yscale('log')
pyplot.xlabel('k')
pyplot.ylabel('P(k)')
#pyplot.show()
title = 'out_degree_' + name
pyplot.savefig(title)
pyplot.clf()
#BA graph w/ same # of vs; show that coefficient is the same
vs = graph.vertices()
bag = BADirectedGraph(5)
bag.build_graph(len(vs)-5)
ins, outs, totals = [], [], []
for vertex in bag.vertices():
deg_in = bag.in_degree(vertex)
deg_out = bag.out_degree(vertex)
deg_tot = deg_in + deg_out
ins.append(deg_in)
outs.append(deg_out)
totals.append(deg_tot)
pmf = Pmf.MakePmfFromList(outs)
xs, ys = pmf.Render()
#convert to log, so we can find line of best fit
xs_log = []
ys_log = []
for x in xs:
if x <= 0:
xs_log.append(.00001)
else:
xs_log.append(math.log(x))
for y in ys:
if y <= 0:
ys_log.append(.00001)
else:
ys_log.append(math.log(y))
coefs = numpy.lib.polyfit(xs_log, ys_log, 1)
fit_y = numpy.lib.polyval(coefs, xs_log)
print coefs
#transform fit line, to plot it on log-log scale
fit_y_log = [math.exp(1) ** f for f in fit_y]
pyplot.plot(xs, ys, 'o')
pyplot.plot(xs, fit_y_log,'r--',linewidth=4)
pyplot.xscale('log')
pyplot.yscale('log')
pyplot.xlabel('k')
pyplot.ylabel('P(k)')
pyplot.show()
title = 'out_degree_BA_on_' + str(len(vs)) +'_vertices'
pyplot.savefig(title)
