1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
|
from analyze import sd_users, build_graph, DATASETS, SYNTH_DATASETS
import matplotlib.pyplot as plt
from matplotlib import rcParams, cm
from matplotlib.colors import Normalize
from matplotlib.pyplot import plot, legend, savefig, xlabel, ylabel,\
hist, title, subplot, tight_layout, ticklabel_format, xlim, ylim
from mpl_toolkits.mplot3d import Axes3D
import numpy as np
import itertools
mq = lambda x: x * 4
def plot_degree_distributions():
plt.figure(figsize=(7, 3))
graph, degrees = build_graph("kiva")
fd_degrees = list(degrees[user] for user in graph)
sd_degrees = list(degrees[user] for user in sd_users(graph))
n, bins, patches = plt.hist(fd_degrees, bins=50, cumulative=True,
label="Initial users", normed=True,
alpha=0.5, histtype="stepfilled")
n, bins, patches = plt.hist(sd_degrees, bins=50, cumulative=True,
histtype="stepfilled", normed=True, alpha=0.5,
label="Friends")
ylim(ymax=1.1)
plt.xlabel("Degree")
plt.ylabel("Probability")
plt.legend(loc="lower right")
plt.savefig("dist.pdf")
def plot_all_performances():
plt.figure(figsize=(7, 14))
for i, dataset in enumerate(DATASETS):
values = [map(float, line.strip().split("\t"))
for line in open(dataset + "_performance.txt")]
a, im, rd, rdf, aps = zip(*values)
a, im, rd, rdf, aps = [map(mq, l) for l in (a, im, rd, rdf, aps)]
a = np.arange(0, 1.001, 0.1)
ax = plt.subplot(5, 2, i + 1)
#ax.set_yscale("log")
plt.plot(a, im, label="Max deg.")
plt.plot(a, rd, label="Rand.")
plt.plot(a, rdf, label="Rand. friend")
plt.plot(a, aps, label="Adapt. Seeding")
plt.xlabel("Budget (fraction of the total number of users)")
plt.ylabel("Performance")
if dataset == "sw":
titl = "SmallWord"
if dataset == "coachella":
titl = "Conf. Model"
if dataset == "kk":
titl = "Kronecker"
if dataset == "b-a":
titl = "Barabasi-Albert"
plt.title(titl)
xlim(xmax=1.1)
plt.legend(loc="upper center", ncol=4, bbox_to_anchor=(0, 0, 1, 1.03),
bbox_transform=plt.gcf().transFigure)
plt.tight_layout()
plt.savefig("test2.pdf")
def compare_performance(fn):
plots = {}
plt.figure()
for dataset in DATASETS:
values = [map(float, line.strip().split("\t"))
for line in open(dataset + "_performance.txt")]
a, im, rd, rdf, aps = zip(*values)
plots[dataset] = [j * 1. / i for (j, i) in zip(aps, im)[1:]]
a = map(mq, a)
for dataset in DATASETS:
plt.plot(a[1:], plots[dataset], label=dataset)
xlim(xmax=550)
plt.xlabel("Budget")
plt.ylabel("Performance")
plt.legend(loc="lower right", ncol=2, fontsize="small")
plt.savefig(fn)
def compare_performance2(fn):
plots = {}
plt.figure()
for dataset in DATASETS:
values = [map(float, line.strip().split("\t"))
for line in open(dataset + "_performance.txt")]
a, im, rd, rdf, aps = zip(*values)
plots[dataset] = [j * 1. / i for (j, i) in zip(aps, im)[1:]]
a = map(mq, a)
a = map(int, a)
z = zip(*plots.itervalues())
means = [np.mean(w) for w in z]
maxi = [np.max(w) for w in z]
mini = [np.min(w) for w in z]
ind = range(len(a[1:]))
width = 0.35
plt.bar(ind, means, width, linewidth=0.1)
plt.errorbar([i + width / 2. for i in ind], means, [mini, maxi], elinewidth=1.2, fmt="none")
plt.xticks([i + width / 2. for i in ind], a[1:])
plt.xlim(-width, len(ind) - 1 + 2 * width)
plt.xlabel("Budget")
plt.ylabel("Relative improvement")
plt.savefig(fn)
def compare_dist():
fd, sd = [], []
plt.figure(figsize=(5, 3))
cm = iter(rcParams["axes.color_cycle"])
for dataset in DATASETS:
graph, degrees = build_graph(dataset)
fd_degrees = list(degrees[user] for user in graph)
sd_degrees = list(degrees[user] for user in sd_users(graph))
fd.append(np.mean(fd_degrees))
sd.append(np.mean(sd_degrees))
ind = range(len(DATASETS))
width = 0.35
plt.bar(ind, fd, width, label="Initial users", color=next(cm))
plt.bar([i + width for i in ind], sd, width, label="Friends",
color=next(cm))
plt.xlim(-width, len(ind) - 1 + 3 * width)
plt.xticks([i + width for i in ind], DATASETS)
plt.ylabel("Avg. degree")
plt.legend()
plt.savefig("para.pdf")
def plot_perf_prob():
plt.figure()
with open("peet_performance_p.txt") as f:
values = [map(float, line.strip().split("\t")) for line in f]
values = zip(*values)
a = [0.01, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]
for i in [0, 1, 2, 3, 5, 9]:
plt.plot(values[0], values[i + 1], label="$p = " + str(a[i]) + "$")
plt.legend()
with open("peet_performance.txt") as f:
values = [map(float, line.strip().split("\t")) for line in f]
values = zip(*values)
plt.gca().set_yscale("log")
plt.xlabel("Budget")
plt.ylabel("Performance")
plt.plot(values[0], values[1], label="Max. degree")
plt.legend(loc="lower right", fontsize="small", ncol=2)
xlim(xmax=450)
plt.savefig("prob.pdf")
def plot_hbo_likes():
plt.figure()
rcParams["font.size"] = 6
with open("hbo_likes_performance.txt") as f:
values = [map(float, line.strip().split("\t")) for line in f]
a, im, aps, apso = zip(*values)
a = np.arange(0, 1.001, 0.1)
plt.gca().set_yscale("log")
#plt.ticklabel_format(style='sci', axis='y', scilimits=(0, 0))
plt.plot(a, map(mq, im), label="Max. degr.")
plt.plot(a, map(mq, aps), label="Adapt. seed. (rest.)")
plt.plot(a, map(mq, apso), label="Adapt. seed.")
plt.xlabel("Budget")
plt.ylabel("Performance")
xlim(xmax=1.1)
plt.legend(loc="lower right")
plt.savefig("hbo_likes.pdf")
def plot_3d():
for dist in ["beta", "gauss"]:
fig = plt.figure()
with open("coachella_performance_p_" + dist + ".txt") as f:
values = [map(float, line.strip().split("\t")) for line in f]
k = np.arange(0, 1.001, 0.1)
ps = np.arange(0.01, 0.99, 0.1)
x, y = np.meshgrid(k, ps)
perfs = [value[1:] for value in values]
perfs = zip(*perfs)
ax = fig.add_subplot(111, projection='3d')
ax.plot_wireframe(x, y, perfs, linewidth=0.1)
ticklabel_format(style='sci', axis='z', scilimits=(0, 0))
xlabel("Budget (fraction of nodes)")
ylabel("Distribution mean")
ax.set_zlabel("Performance")
ax.invert_xaxis()
plt.savefig(dist + ".pdf")
plt.show()
def plot_time():
plt.figure()
rcParams["font.size"] = 6
a1 = np.loadtxt("time_aps_100.txt")
a2 = np.loadtxt("time_aps_500.txt")
lp1 = np.loadtxt("time_lp_100.txt")
lp2 = np.loadtxt("time_lp_500.txt")
subplot(2, 2, 1)
plot(a1[:, 0], a1[:, 1], "-", label="Comb.")
plot(lp1[:, 0], lp1[:, 1], "-", label="LP")
xlabel("n")
ylabel("time (s)")
xlim(0, 100000)
legend(loc="upper left")
ticklabel_format(style='sci', axis='x', scilimits=(0, 0))
subplot(2, 2, 2)
plot(a1[:, 0], a1[:, 2], "-", label="Comb.")
plot(lp1[:, 0], lp1[:, 2], "-", label="LP")
ticklabel_format(style='sci', axis='x', scilimits=(0, 0))
xlabel("n")
ylabel("\# cycles")
xlim(0, 100000)
legend(loc="upper left")
subplot(2, 2, 3)
plot(a2[:, 0], a2[:, 1], "-", label="Comb.")
plot(lp2[:, 0], lp2[:, 1], "-", label="LP")
ticklabel_format(style='sci', axis='x', scilimits=(0, 0))
xlabel("n")
ylabel("time (s)")
xlim(0, 100000)
legend(loc="upper left")
subplot(2, 2, 4)
plot(a2[:, 0], a2[:, 2], "-", label="Comb.")
plot(lp2[:, 0], lp2[:, 2], "-", label="LP")
ticklabel_format(style='sci', axis='x', scilimits=(0, 0))
xlabel("n")
ylabel("\# cycles")
xlim(0, 100000)
legend(loc="upper left")
tight_layout(h_pad=-0.5)
savefig("time.pdf")
if __name__ == "__main__":
SYNTH_DATASETS = ["b-a", "kk", "sw", "coachella"]
DATASETS = SYNTH_DATASETS
plot_all_performances()
#plot_3d()
#plot_hbo_likes()
#compare_performance()
#plot_perf_prob()
#compare_dist()
#plot_time()
#plot_degree_distributions()
#for style in plt.style.available:
# plt.style.use(style)
# compare_performance("performance_" + style + ".pdf")
#compare_performance2("comp4_" + ".pdf")
|
