def plot(cluster,
funcs,
labels,
filename="fingerprints.pdf",
plot_title='Scores by Cluster',
cmap='Paired',
filter=lambda a: True,
plot_pop=False,
run_dir="../run/"):
assert len(labels) == len(funcs), "must have a label for every func"
# load in cluster data
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
# compress clusters
# TODO: replace with call to agent.cluster.compress(700)
clusters = [[i, clust] for i, clust in enumerate(clusters)
if len(clust) > 700]
ids, clusters = zip(*clusters)
num_clusters = len(clusters)
# build up cluster data
data = [[
np.average([func(a) for a in clust if filter(a)]) for clust in clusters
] for func in funcs]
err = [[
np.std([func(a) for a in clust if filter(a)]) /
np.sqrt(len([func(a) for a in clust if filter(a)]))
for clust in clusters
] for func in funcs]
# plot actual bars for each function
ind = np.arange(num_clusters)
width = 1.0 / (len(funcs) + 1)
bars = []
for offset, datum in enumerate(data):
b = bar(ind + (offset * width),
datum,
width,
color=cm.Paired(float(offset) / len(funcs)),
yerr=err[offset],
capsize=1.5)
bars.append(b)
# generate final plot
title(plot_title, weight='black')
ylabel('Normalized Value', weight='bold')
ylim(ymin=0, ymax=1.0)
xlabel('Cluster', weight='bold')
xticks(ind + (width * len(funcs) / 2),
["%d" % ids[i] for i, clust in enumerate(clusters)])
legend([b[0] for b in bars], labels, loc='upper left')
savefig(filename, dpi=200)
def plot(cluster_file,
run_dir='../run/',
anim_path=None,
frame_path=None,
min_size=None):
# configure default animation and frame path
if anim_path is None:
anim_path = 'anim'
if frame_path is None:
frame_path = os.path.join(anim_path, 'frames')
# ensure paths exist
if not os.path.isdir(anim_path):
os.mkdir(anim_path)
if not os.path.isdir(frame_path):
os.mkdir(frame_path)
# load clusters
clusters = rc.load_clusters(cluster_file, sort='random')
# compress, if a threshold is set
if min_size is not None:
clusters = compress_clusters(clusters, min_size)
# calculate number of clusters
n_clusters = len(clusters)
agent_cluster = rc.agent_cluster(clusters)
# establish cluster object as a shared memory object.
#manager = Manager()
#cluster = manager.dict(agent_cluster)
start = 1
stop = 300
# grouping to keep memory footprint reasonably small.
# TODO: Fix this using iterators and multiprocessing.Pool(). Running into
# major issues with serialization of Agent class that is preventing
# plot_step from being picklable.
for begin in xrange(start, stop, 1500):
end = min(begin + 1499, stop)
p = get_population_during_time(begin, end)
for t in xrange(begin, end, 3):
plot_step(t,
p,
agent_cluster,
n_clusters,
frame_path,
cmap='Paired')
subprocess.call((
"mencoder mf://%s/*.png -o %s/output.avi -mf type=png:w=600:h=800:fps=30 -ovc x264 -x264encopts qp=20"
% (frame_path, anim_path)).split())
def plot(cluster, filename="plot.png", func=lambda a: a.id,
plot_title='', cmap='Paired', filter=lambda a: True,
draw_legend=False, radius='2.25', sym=None):
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
p = get_population()
pops = [0 for i in range(30000)]
cluster_pops = []
cluster_pop_max = []
for clust in range(len(clusters)):
cluster_pops.append(pops[:])
cluster_pop_max.append([0,0,-1,0])
for clust,agents in enumerate(clusters):
for agent in agents:
a = Agent(agent)
for i in range(a.birth, a.death):
cluster_pops[clust][i] += 1
if cluster_pop_max[clust][2] == -1:
cluster_pop_max[clust][2] = i
if i > cluster_pop_max[clust][3]:
cluster_pop_max[clust][3] = i
if cluster_pops[clust][i] > cluster_pop_max[clust][0]:
cluster_pop_max[clust][0] = cluster_pops[clust][i]
cluster_pop_max[clust][1] = i
lines=[]
for i,clust in enumerate(cluster_pops):
lines.append(pylab.plot(range(30000),clust,
label=("%d: k=%d" % (i, len(clusters[i]))),
color=pylab.cm.Paired(float(i)/len(clusters))))
if draw_legend:
pylab.figlegend(lines, ["%d: k=%d" % (i, len(clust))
for i,clust in enumerate(clusters)],
'center right',
ncol=((len(clusters)/35)+1), prop=dict(size=6))
else:
print "should not draw!!!"
title = r"Cluster Population ($\epsilon$ = %s, %d clusters)" % (radius, len(clusters))
pylab.title(title, weight='black')
pylab.xlabel("Time", weight='bold')
pylab.ylabel("Population Size", weight='bold')
if sym is not None:
pylab.figtext(0,.954, '(%s)' % sym, size=6, weight='black')
pylab.savefig(filename, dpi=300)
print 'cluster, totalPop, start, peak, stop, maxPop'
for clust,agents in enumerate(clusters):
print clust, len(agents), cluster_pop_max[clust][2], cluster_pop_max[clust][1], cluster_pop_max[clust][3]+1, cluster_pop_max[clust][0]
def plot(cluster_file,
filename="plot.png",
func=lambda a: a.id,
plot_title='',
cmap='Paired',
filter=lambda a: True,
draw_legend=False,
radius='2.25',
sym=None,
run_dir='../run/'):
"""
Creates a line plot showing population per cluster
"""
# retrieve cluster data from cluster file
clusters = rc.load_clusters(cluster_file)
# grab cluster population
p = get_population(run_dir=run_dir)
lines = []
for cluster, agents in enumerate(clusters):
pop_by_time = list(repeat(0, 30000))
for agent in agents:
a = Agent(agent)
for i in range(a.birth, a.death):
pop_by_time[i] += 1
lines.append(
pylab.plot(range(30000),
pop_by_time,
label=("%d: k=%d" % (i, len(clusters[cluster]))),
color=pylab.cm.Paired(float(cluster) / len(clusters))))
if draw_legend:
pylab.figlegend(lines, [
"%d: k=%d" % (i, len(cluster))
for i, cluster in enumerate(clusters)
],
'center right',
ncol=((len(clusters) / 35) + 1),
prop=dict(size=6))
title = r"Cluster Population ($\epsilon$ = %s, %d clusters)" % (
radius, len(clusters))
pylab.title(title, weight='black')
pylab.xlabel("Time", weight='bold')
pylab.ylabel("Population Size", weight='bold')
if sym is not None:
pylab.figtext(0, .954, '(%s)' % sym, size=6, weight='black')
pylab.savefig(filename, dpi=300)
def plot(cluster, filename="plot.png", plot_title='', cmap='RdBu_r'):
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
p = get_population()[400:4000]
p = [a for a in p if a.birth != 30000]
#normalize genomes
print "normalizing genomes..."
gs = zscore([a.genome[:] for a in p], axis=1)
for i, agent in enumerate(p):
agent.ngenome = gs[i]
del agent.genome
print "calculating distances..."
gene_results = []
geo_results = []
pool = Pool()
for i,agent1 in enumerate(p):
print agent1.id, "to everything else"
x = [(agent1.ngenome, agent2.ngenome) for agent2 in p[i+1:]]
r = pool.map(dists_wrapper, x)
gene_results.extend(r)
y = [(agent1.positions[agent1.birth+1],
agent2.positions[agent2.birth+1])
for agent2 in p[i+1:]]
r = pool.map(dists_wrapper, y)
geo_results.extend(r)
#for agent2 in p[i+1:]:
# results.append(dists(agent1, agent2))
del agent1.ngenome
del agent1.positions
xs = geo_results
ys = gene_results
print "plotting..."
hexbin(xs, ys, cmap=cmap, alpha=0.6, edgecolors='none', mincnt=25)
title(plot_title)
xlabel('Geographic Distance')
ylabel('Genetic Distance')
xlim( (0,142) )
ylim( (1000,6000) )
colorbar()
savefig(filename, dpi=200)
def plot(cluster, filename="plot.png", plot_title='', cmap='RdBu_r'):
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
p = get_population()[400:4000]
p = [a for a in p if a.birth != 30000]
#normalize genomes
print "normalizing genomes..."
gs = zscore([a.genome[:] for a in p], axis=1)
for i, agent in enumerate(p):
agent.ngenome = gs[i]
del agent.genome
print "calculating distances..."
gene_results = []
geo_results = []
pool = Pool()
for i, agent1 in enumerate(p):
print agent1.id, "to everything else"
x = [(agent1.ngenome, agent2.ngenome) for agent2 in p[i + 1:]]
r = pool.map(dists_wrapper, x)
gene_results.extend(r)
y = [(agent1.positions[agent1.birth + 1],
agent2.positions[agent2.birth + 1]) for agent2 in p[i + 1:]]
r = pool.map(dists_wrapper, y)
geo_results.extend(r)
#for agent2 in p[i+1:]:
# results.append(dists(agent1, agent2))
del agent1.ngenome
del agent1.positions
xs = geo_results
ys = gene_results
print "plotting..."
hexbin(xs, ys, cmap=cmap, alpha=0.6, edgecolors='none', mincnt=25)
title(plot_title)
xlabel('Geographic Distance')
ylabel('Genetic Distance')
xlim((0, 142))
ylim((1000, 6000))
colorbar()
savefig(filename, dpi=200)
def plot(cluster,
filename="plot-filter.eps",
funcs=(lambda a: a.id),
plot_title='Scores by Cluster',
cmap='Paired',
filter=lambda a: True,
labels=('id'),
plot_pop=False):
assert len(labels) == len(funcs), "must have a label for every func"
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
clusters = [[i, clust] for i, clust in enumerate(clusters)
if len(clust) > 700]
ids, clusters = zip(*clusters)
num_clusters = len(clusters)
ind = np.arange(num_clusters)
if plot_pop:
width = 1.0 / (len(funcs) + 2)
else:
width = 1.0 / (len(funcs) + 1)
data = [[
np.average([func(a) for a in clust if filter(a)]) for clust in clusters
] for func in funcs]
if plot_pop:
data.append([len(clust) / 25346.0 for clust in clusters])
labels.append('population')
bars = []
for offset, datum in enumerate(data):
b = bar(ind + (offset * width),
datum,
width,
color=cm.Paired(float(offset) / len(funcs)))
bars.append(b)
ylabel('normalized value')
title(plot_title)
ylim(ymin=0)
xticks(ind + (width * len(funcs) / 2), [
"%d: k=%d" % (ids[i], len(clust)) for i, clust in enumerate(clusters)
],
fontsize=8)
legend([b[0] for b in bars], labels, loc='upper left', prop=dict(size=6))
savefig(filename, dpi=200)
def plot(cluster, funcs, labels, filename="fingerprints.pdf",
plot_title='Scores by Cluster', cmap='Paired', filter=lambda a: True,
plot_pop=False, run_dir="../run/"):
assert len(labels) == len(funcs), "must have a label for every func"
# load in cluster data
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
# compress clusters
# TODO: replace with call to agent.cluster.compress(700)
clusters = [[i, clust] for i,clust in enumerate(clusters)
if len(clust) > 700]
ids, clusters = zip(*clusters)
num_clusters = len(clusters)
# build up cluster data
data = [[np.average([func(a) for a in clust if filter(a)])
for clust in clusters]
for func in funcs]
err = [[np.std([func(a) for a in clust if filter(a)]) /
np.sqrt(len([func(a) for a in clust if filter(a)]))
for clust in clusters]
for func in funcs]
# plot actual bars for each function
ind = np.arange(num_clusters)
width = 1.0 / (len(funcs) + 1)
bars = []
for offset,datum in enumerate(data):
b = bar(ind+(offset*width), datum, width,
color=cm.Paired(float(offset)/len(funcs)),
yerr=err[offset], capsize=1.5)
bars.append(b)
# generate final plot
title(plot_title, weight='black')
ylabel('Normalized Value', weight='bold')
ylim(ymin=0,ymax=1.0)
xlabel('Cluster', weight='bold')
xticks(ind + ( width*len(funcs) / 2),
["%d" % ids[i]
for i, clust in enumerate(clusters)])
legend([b[0] for b in bars], labels, loc='upper left')
savefig(filename, dpi=200)
def plot(cluster, filename="plot.png", func=lambda a: a.id,
plot_title='', cmap='Paired', filter=lambda a: True):
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
p = get_population()
p.sort(key=lambda a:-len(clusters[ac[a.id]]))
xs = [(a.birth + a.death) / 2 for a in p if filter(a)]
ys = [func(a) for a in p if filter(a)]
cs = [ac[a.id] for a in p if filter(a)]
scatter(xs,ys, cmap=cmap, c=cs, s=10, alpha=0.6, edgecolors='none')
title(plot_title)
xlim(0,30000)
ylim(ymin=0)
savefig(filename, dpi=200)
def plot(cluster_file, run_dir='../run/', anim_path=None, frame_path=None,
min_size=None):
# configure default animation and frame path
if anim_path is None:
anim_path = 'anim'
if frame_path is None:
frame_path = os.path.join(anim_path, 'frames')
# ensure paths exist
if not os.path.isdir(anim_path):
os.mkdir(anim_path)
if not os.path.isdir(frame_path):
os.mkdir(frame_path)
# load clusters
clusters = rc.load_clusters(cluster_file, sort='random')
# compress, if a threshold is set
if min_size is not None:
clusters = compress_clusters(clusters, min_size)
# calculate number of clusters
n_clusters = len(clusters)
agent_cluster = rc.agent_cluster(clusters)
# establish cluster object as a shared memory object.
#manager = Manager()
#cluster = manager.dict(agent_cluster)
start = 1
stop = 300
# grouping to keep memory footprint reasonably small.
# TODO: Fix this using iterators and multiprocessing.Pool(). Running into
# major issues with serialization of Agent class that is preventing
# plot_step from being picklable.
for begin in xrange(start, stop, 1500):
end = min(begin + 1499, stop)
p = get_population_during_time(begin, end)
for t in xrange(begin, end, 3):
plot_step(t, p, agent_cluster, n_clusters, frame_path, cmap='Paired')
subprocess.call(("mencoder mf://%s/*.png -o %s/output.avi -mf type=png:w=600:h=800:fps=30 -ovc x264 -x264encopts qp=20" % (frame_path, anim_path) ).split())
def plot(cluster,
filename="plot.png",
func=lambda a: a.id,
plot_title='',
cmap='Paired',
filter=lambda a: True):
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
p = get_population()
p.sort(key=lambda a: -len(clusters[ac[a.id]]))
xs = [(a.birth + a.death) / 2 for a in p if filter(a)]
ys = [func(a) for a in p if filter(a)]
cs = [ac[a.id] for a in p if filter(a)]
scatter(xs, ys, cmap=cmap, c=cs, s=10, alpha=0.6, edgecolors='none')
title(plot_title)
xlim(0, 30000)
ylim(ymin=0)
savefig(filename, dpi=200)
def plot(cluster_file, filename="plot.png", func=lambda a: a.id,
plot_title='', cmap='Paired', filter=lambda a: True,
draw_legend=False, radius='2.25', sym=None, run_dir='../run/'):
"""
Creates a line plot showing population per cluster
"""
# retrieve cluster data from cluster file
clusters = rc.load_clusters(cluster_file)
# grab cluster population
p = get_population(run_dir=run_dir)
lines=[]
for cluster, agents in enumerate(clusters):
pop_by_time = list(repeat(0, 30000))
for agent in agents:
a = Agent(agent)
for i in range(a.birth, a.death):
pop_by_time[i] += 1
lines.append(
pylab.plot(range(30000), pop_by_time,
label=("%d: k=%d" % (i, len(clusters[cluster]))),
color=pylab.cm.Paired(float(cluster)/len(clusters))))
if draw_legend:
pylab.figlegend(lines, ["%d: k=%d" % (i, len(cluster))
for i,cluster in enumerate(clusters)],
'center right',
ncol=((len(clusters)/35)+1), prop=dict(size=6))
title = r"Cluster Population ($\epsilon$ = %s, %d clusters)" % (radius, len(clusters))
pylab.title(title, weight='black')
pylab.xlabel("Time", weight='bold')
pylab.ylabel("Population Size", weight='bold')
if sym is not None:
pylab.figtext(0,.954, '(%s)' % sym, size=6, weight='black')
pylab.savefig(filename, dpi=300)
def plot(cluster, filename="plot-filter.eps", funcs=(lambda a: a.id),
plot_title='Scores by Cluster', cmap='Paired', filter=lambda a: True,
labels=('id'), plot_pop=False):
assert len(labels) == len(funcs), "must have a label for every func"
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
clusters = [[i, clust] for i,clust in enumerate(clusters)
if len(clust) > 700]
ids, clusters = zip(*clusters)
num_clusters = len(clusters)
ind = np.arange(num_clusters)
if plot_pop:
width = 1.0 / (len(funcs) + 2)
else:
width = 1.0 / (len(funcs) + 1)
data = [[np.average([func(a) for a in clust if filter(a)])
for clust in clusters]
for func in funcs]
if plot_pop:
data.append([len(clust) / 25346.0 for clust in clusters])
labels.append('population')
bars = []
for offset,datum in enumerate(data):
b = bar(ind+(offset*width), datum, width,
color=cm.Paired(float(offset)/len(funcs)))
bars.append(b)
ylabel('normalized value')
title(plot_title)
ylim(ymin=0)
xticks(ind + ( width*len(funcs) / 2),
["%d: k=%d" % (ids[i], len(clust))
for i, clust in enumerate(clusters)],
fontsize=8)
legend([b[0] for b in bars], labels, loc='upper left', prop=dict(size=6))
savefig(filename, dpi=200)
def plot(cluster,
filename="plot.png",
func=lambda a: a.id,
plot_title='',
cmap='Paired',
filter=lambda a: True,
draw_legend=False,
radius='2.25',
sym=None):
ac = rc.load(cluster)
clusters = rc.load_clusters(cluster)
p = get_population()
pops = [0 for i in range(30000)]
cluster_pops = []
cluster_pop_max = []
for clust in range(len(clusters)):
cluster_pops.append(pops[:])
cluster_pop_max.append([0, 0, -1, 0])
for clust, agents in enumerate(clusters):
for agent in agents:
a = Agent(agent)
for i in range(a.birth, a.death):
cluster_pops[clust][i] += 1
if cluster_pop_max[clust][2] == -1:
cluster_pop_max[clust][2] = i
if i > cluster_pop_max[clust][3]:
cluster_pop_max[clust][3] = i
if cluster_pops[clust][i] > cluster_pop_max[clust][0]:
cluster_pop_max[clust][0] = cluster_pops[clust][i]
cluster_pop_max[clust][1] = i
lines = []
for i, clust in enumerate(cluster_pops):
lines.append(
pylab.plot(range(30000),
clust,
label=("%d: k=%d" % (i, len(clusters[i]))),
color=pylab.cm.Paired(float(i) / len(clusters))))
if draw_legend:
pylab.figlegend(
lines,
["%d: k=%d" % (i, len(clust)) for i, clust in enumerate(clusters)],
'center right',
ncol=((len(clusters) / 35) + 1),
prop=dict(size=6))
else:
print "should not draw!!!"
title = r"Cluster Population ($\epsilon$ = %s, %d clusters)" % (
radius, len(clusters))
pylab.title(title, weight='black')
pylab.xlabel("Time", weight='bold')
pylab.ylabel("Population Size", weight='bold')
if sym is not None:
pylab.figtext(0, .954, '(%s)' % sym, size=6, weight='black')
pylab.savefig(filename, dpi=300)
print 'cluster, totalPop, start, peak, stop, maxPop'
for clust, agents in enumerate(clusters):
print clust, len(agents), cluster_pop_max[clust][2], cluster_pop_max[
clust][1], cluster_pop_max[clust][3] + 1, cluster_pop_max[clust][0]
os.mkdir(frame_path)
'''
print "unpickling clustsers ..."
c = Cluster.load(sys.argv[-1])
print "sorting clustsers ..."
c.sort()
cluster = c.agents
del c
print cluster
exit()
'''
import readcluster
clusters = readcluster.load_clusters(sys.argv[-1])
new_clust = [clust for clust in clusters if len(clust) > 700]
small_clust = []
for clust in clusters:
if len(clust) <= 700:
small_clust.extend(clust)
new_clust.append(small_clust)
cluster = readcluster.agent_cluster(new_clust)
n_clusters = len(set(cluster.values()))
print n_clusters
cluster_pops[clust][i] += 1
# set cluster start
if cluster_pop_max[clust][2] == -1:
cluster_pop_max[clust][2] = i
# set cluster stop
if i > cluster_pop_max[clust][3]:
cluster_pop_max[clust][3] = i
# set max Population
if cluster_pops[clust][i] > cluster_pop_max[clust][0]:
cluster_pop_max[clust][0] = cluster_pops[clust][i]
cluster_pop_max[clust][1] = i
print 'cluster, totalPop, start, peak, stop, maxPop'
for clust, agents in enumerate(clusters):
print clust, len(agents), cluster_pop_max[clust][2], cluster_pop_max[
clust][1], cluster_pop_max[clust][3] + 1, cluster_pop_max[clust][0]
if __name__ == '__main__':
import sys
# get clusters
cluster_file = sys.argv[-1]
ac = rc.load(cluster_file)
clusters = rc.load_clusters(cluster_file)
print_timeline(clusters)
os.mkdir(frame_path)
'''
print "unpickling clustsers ..."
c = Cluster.load(sys.argv[-1])
print "sorting clustsers ..."
c.sort()
cluster = c.agents
del c
print cluster
exit()
'''
import readcluster
clusters = readcluster.load_clusters(sys.argv[-1])
new_clust = [clust for clust in clusters if len(clust) > 700]
small_clust = []
for clust in clusters:
if len(clust) <= 700:
small_clust.extend(clust)
new_clust.append(small_clust)
cluster = readcluster.agent_cluster(new_clust)
n_clusters = len(set(cluster.values()))
print n_clusters
print "processing..."
a = Agent(agent)
for i in range(a.birth, a.death):
cluster_pops[clust][i] += 1
# set cluster start
if cluster_pop_max[clust][2] == -1:
cluster_pop_max[clust][2] = i
# set cluster stop
if i > cluster_pop_max[clust][3]:
cluster_pop_max[clust][3] = i
# set max Population
if cluster_pops[clust][i] > cluster_pop_max[clust][0]:
cluster_pop_max[clust][0] = cluster_pops[clust][i]
cluster_pop_max[clust][1] = i
print 'cluster, totalPop, start, peak, stop, maxPop'
for clust,agents in enumerate(clusters):
print clust, len(agents), cluster_pop_max[clust][2], cluster_pop_max[clust][1], cluster_pop_max[clust][3]+1, cluster_pop_max[clust][0]
if __name__ == '__main__':
import sys
# get clusters
cluster_file = sys.argv[-1]
ac = rc.load(cluster_file)
clusters = rc.load_clusters(cluster_file)
print_timeline(clusters)
