-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathplot_results.py
559 lines (479 loc) · 25.6 KB
/
plot_results.py
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
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
import os
import numpy as np
import pandas as pd
import matplotlib
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import seaborn as sns; sns.set(style="whitegrid")
from PIL import Image
from tqdm import tqdm
##############################################################################
## Load Data ############################################################################
# old_rows = {'noise':[],'p_in':[],'mult':[],'instance':[],'repetition':[],'algorithm':[],'accuracy':[],'robustness':[],'seconds':[],'method':[]}
# print('entrou')
# print(df.columns)
# print(df2.columns)
# ['noise', 'mult', 'p_in', 'p_out', 'instance', 'repetition',
# 'hedonic_accuracy', 'hedonic_infos', 'hedonic_robustness',
# 'hedonic_time', 'local improve_accuracy', 'local improve_infos',
# 'local improve_robustness', 'local improve_time', 'spectral_accuracy',
# 'spectral_infos', 'spectral_robustness', 'spectral_time', 'algorithm',
# 'accuracy', 'robustness', 'method', 'seconds'],
# ['p_in', 'mult', 'instance', 'repetition', 'algorithm', 'accuracy',
# 'robustness', 'method', 'seconds', 'noise'],
# print(df3.columns)
# print(df2.columns)
# print(df3.head(10))
# df3 = df.melt('mult', var_name='algorithm', value_name='noise')
# for index, row in tqdm(df.iterrows()):
# print(index, len(df), index/len(df)*100)
# for alg in ['hedonic', 'local improve', 'spectral']:
# pass
# old_rows['noise'].append(row['noise'])
# old_rows['p_in'].append(row['p_in'])
# old_rows['mult'].append(row['mult'])
# old_rows['instance'].append(row['instance'])
# old_rows['repetition'].append(row['repetition'])
# old_rows['algorithm'].append(alg)
# old_rows['accuracy'].append(row[f'{alg}_accuracy'])
# old_rows['robustness'].append(row[f'{alg}_robustness'])
# old_rows['seconds'].append(row[f'{alg}_time'])
# old_rows['method'].append('dist')
# df2 = df2.append({
# 'noise': row['noise'],
# 'p_in': row['p_in'],
# 'mult': row['mult'],
# 'instance': row['instance'],
# 'repetition': row['repetition'],
# 'algorithm': alg,
# 'accuracy': row[f'{alg}_accuracy'],
# 'robustness': row[f'{alg}_robustness'],
# 'seconds': row[f'{alg}_time'],
# 'method': 'dist'
# }, ignore_index=True)
# clustering / improvement / robust / local improve == local improve
algorithms = { 'spectral':'spectral', 'local improve':'local improve', 'hedonic':'hedonic', 'ml':'Louvain', 'ecg':'ECG' }
alg_colors = { 'spectral':'#C44E53', 'local improve':'#DD8452', 'hedonic':'#55A869', 'Louvain':'#4C72B0', 'ECG':'#8172B3' }
file_name_prefix = ''
def adaptd_df_to_new_exp(df, prefix='outputs/noises/'):
df1 = pd.read_csv(f'{prefix}ps=10_mults=11_inst=10_reps=10_nComm=2_commSize=500.csv')
df1 = df1.loc[df1['method'] == 'dist']
df1 = df1.loc[df1['algorithm'] != 'hedonic']
df1 = df1.loc[df1['algorithm'] != 'spectral']
df1.drop(columns=['p_in','instance','repetition','robustness','method'], inplace=True)
df2, temp = pd.DataFrame(), df1.copy()
for noise in df['noise'].unique():
temp['noise'] = [noise] * len(temp['mult'])
df2 = pd.concat([df2, temp])
fix_cols = ['noise', 'p_in', 'mult', 'instance', 'repetition']
dfs_melted = {}
for col in ['accuracy', 'time', 'robustness']:
dfs_melted[col] = df.melt(id_vars=fix_cols, value_vars=[f'hedonic_{col}', f'onepass_{col}', f'spectral_{col}'], var_name='algorithm', value_name=col)
df3 = dfs_melted['accuracy'].loc[:, fix_cols+['algorithm']]
df3['method'] = ['dist'] * len(df3)
df3['algorithm'] = [alg.split('_')[0] for alg in df3['algorithm']]
df3['accuracy'] = dfs_melted['accuracy']['accuracy']
df3['robustness'] = dfs_melted['robustness']['robustness']
df3['seconds'] = dfs_melted['time']['time']
df2 = pd.concat([df2, df3])
df2['algorithm'] = ['local improve' if alg == 'onepass' else algorithms[alg] for alg in df2['algorithm'].values]
return df2
# merge old and new dataframes
def load_df():
prefix = 'outputs/noises/'
df = pd.concat([pd.read_csv(f'{prefix}{csv}') for csv in os.listdir(prefix) if csv[-3:] == 'csv' and csv[:5] == 'noise'])
# df['p_in-p_out'] = df['p_in'] - df['p_out']
return adaptd_df_to_new_exp(df, prefix)
# one datafram
# def load_df():
# prefix = 'outputs/recents'
# csv_name = 'with_noises_fix__ps=5_mults=11_inst=5_reps=15_noises=13_nComm=2_commSize=500'
# df = pd.read_csv(f'{prefix}/{csv_name}.csv')
# df = df.loc[df['method'] == 'dist']
# df.drop(columns=['p_in','instance','repetition','robustness','method'], inplace=True)
# df['algorithm'] = [alg.split('_')[0] for alg in df['algorithm']]
# df['algorithm'] = [algorithms[alg] for alg in df['algorithm'].values]
# return df
##############################################################################
def load_dfs_noise(df):
dfs_noise = dict(tuple(df.groupby('noise')))
# went, total = 0, len(dfs_noise) * 11 * 25 * 25 * 3
for noise, data in dfs_noise.items():
df_noise = data.loc[:, ['mult', 'algorithm', 'accuracy']]
# df_noise = pd.DataFrame(columns=['mult', 'algorithm', 'accuracy']) # +[f'{alg}_accuracy' for alg in algorithms]
# for mult in data['mult'].unique():
# # print('noise =', noise, '-- mult =', mult)
# for inst in data['instance'].unique():
# for alg in data['algorithm'].unique(): # [f'{alg}_accuracy' for alg in algorithms]
# went += 1
# print(went, total, went/total*100)
# df_noise = df_noise.append({
# 'mult' : mult,
# 'algorithm': alg,
# 'accuracy': np.mean(data.loc[(data['mult'] == mult) & (data['instance'] == inst) & (data['algorithm'] == alg), 'accuracy'].values),
# # algs[0] : np.mean(data.loc[(data['mult'] == mult) & (data['instance'] == inst), [algs[0]]].values),
# # algs[1] : np.mean(data.loc[(data['mult'] == mult) & (data['instance'] == inst), [algs[1]]].values),
# # algs[2] : np.mean(data.loc[(data['mult'] == mult) & (data['instance'] == inst), [algs[2]]].values),
# }, ignore_index=True)
# df_noise = data.loc[:, ['mult']+[f'{alg}_accuracy' for alg in algorithms]]
# df_noise = df_noise.melt('mult', var_name='algorithm', value_name='accuracy')
# df_noise['algorithm'] = [alg.split('_')[0] for alg in df_noise['algorithm'].values]
# df_noise['algorithm'] = [algorithms[alg] for alg in df_noise['algorithm'].values]
df_noise.rename(columns = {'mult':'q/p'}, inplace = True)
dfs_noise[noise] = df_noise
return dfs_noise
##############################################################################
def load_speed(df):
speed = df.loc[:, ['mult', 'noise', 'algorithm', 'seconds']] # +[f'{alg}_time' for alg in algorithms]
# speed = speed.melt('noise', var_name='algorithm', value_name='seconds')
# speed['algorithm'] = [alg.split('_')[0] for alg in speed['algorithm'].values]
speed['algorithm'] = [alg.replace(' ','\n') for alg in speed['algorithm'].values]
for alg in speed['algorithm'].unique():
alg_df = speed[speed['algorithm'] == alg]
return speed#.loc[speed['algorithm'] != 'ECG']
##############################################################################
def calc_r(df_accs, noise=.5):
mults = df_accs['q/p'].unique()
mults.sort()
accuracies = []
for mult in mults:
accuracies.append(np.mean(df_accs.loc[(df_accs['q/p'] == mult), 'accuracy'].values))
for i, acc in enumerate(accuracies):
if acc < 1-noise:
if i == 0:
return 0
else:
ab = accuracies[i-1] - acc
bc = mults[i] - mults[i-1]
ef = 1 - noise - acc
fc = (bc * ef) / ab
return mults[i] - fc
return 1
def load_df_A(dfs_noise):
# noises = [round(x, 3) for x in [0, 0.025] + list(np.linspace(0, .5, 11))[1:-1] + [0.475, .5]]
df_A = pd.DataFrame(columns=['noise', 'algorithm', 'r'])
for noi in list(dfs_noise):
df_noi = dfs_noise[noi]
for alg in df_noi['algorithm'].unique():
df_A = df_A.append({
'noise' : noi,
'algorithm': alg,
'r': calc_r(df_noi[df_noi['algorithm'] == alg].loc[:,['q/p','accuracy']],noi)
}, ignore_index=True)
df_A['r'] = df_A['r'].apply(pd.to_numeric)
df_A['noise'] = df_A['noise'].apply(pd.to_numeric)
df_A.rename(columns={'r':'r = maximum tolerated q/p'}, inplace=True)
# denom = 1468.49
# df_A['spectral'] = [0,(denom-1028.25)/denom,(denom-900.48)/denom,(denom-710.51)/denom,(denom-600.57)/denom,(denom-536.37)/denom,(denom-474.91)/denom,(denom-414.02)/denom,(denom-370.85)/denom,(denom-325.62)/denom,(denom-245.38)/denom,(denom-170.39)/denom,1]
# df_A['local improve'] = [0,(denom-824.52)/denom,(denom-700.02)/denom,(denom-591.54)/denom,(denom-543.63)/denom,(denom-512.04)/denom,(denom-487.50)/denom,(denom-470.42)/denom,(denom-462.15)/denom,(denom-453.79)/denom,(denom-448.22)/denom,(denom-389.49)/denom,1]
# df_A['hedonic'] = [0,(denom-857.10)/denom,(denom-715.39)/denom,(denom-575.51)/denom,(denom-502.74)/denom,(denom-438.50)/denom,(denom-397.13)/denom,(denom-362.14)/denom,(denom-338.83)/denom,(denom-302.52)/denom,(denom-303.45)/denom,(denom-217.39)/denom,1]
# df_A = df_A.melt('noise', var_name='algorithm', value_name='r = maximum tolerated q/p') # r is the maximum value of q/p for which we observe gains in the corresponding method when contrasted against simple replication of the input as the output, i.e., maximum value of q/p for which the accuracy of the method is greater than 1-noise
# df_A['algorithm'] = [algorithms[alg] for alg in df_A['algorithm'].values]
return df_A
##############################################################################
def load_realnets_df():
prefix = 'outputs/real_nets'
csv_name = 'agora_vai__networks=4_reps=1000_noises=13'
realnets_df = pd.read_csv(f'{prefix}/{csv_name}.csv')
realnets_df = realnets_df.loc[realnets_df['method'] == 'dist']
realnets_df.drop(columns=['repetition','robustness','method'], inplace=True)
# print(realnets_df['algorithm'].unique())
# print(realnets_df.loc[realnets_df['algorithm'] == 'hedonic_n0'].mean())
# print(realnets_df.loc[realnets_df['algorithm'] == 'local improve_n0'].mean())
# print(realnets_df.loc[realnets_df['algorithm'] == 'hedonic_n0'].max())
# print(realnets_df.loc[realnets_df['algorithm'] == 'local improve_n0'].max())
####################
# prefix = 'outputs/real_nets'
# csv_name = '4_networks_10_repts'
# realnets_df = pd.read_csv(f'{prefix}/{csv_name}.csv')
# realnets_df.drop(columns=['infos','robustness'], inplace=True)
# realnets_df.rename(columns={'time':'seconds'}, inplace=True)
# new_realnets_df = pd.read_csv(f'{prefix}/real_nets__networks=4_reps=10.csv')
# new_realnets_df = new_realnets_df.loc[new_realnets_df['method'] == 'dist']
# new_realnets_df.drop(columns=['repetition','robustness','method'], inplace=True)
# new_realnets_df = new_realnets_df.loc[new_realnets_df['algorithm'] != 'hedonic']
# new_realnets_df = new_realnets_df.loc[new_realnets_df['algorithm'] != 'spectral']
# realnets_df = pd.concat([realnets_df, new_realnets_df])
####################
# print(realnets_df['algorithm'].unique())
# algs = []
# for alg in realnets_df['algorithm']:
# if alg != [algorithms['spectral'], algorithms['ml'], algorithms['ECG']]:
# alg = alg.split('_')
# alg[0] = algorithms[alg[0]]
# alg = f'{alg[0]} (noi={alg[1][2:]})'
# algs.append(alg)
# realnets_df['algorithm'] = algs
# print(realnets_df.loc[(realnets_df['algorithm'] == 'ml') & (realnets_df['network'] == 'pol_blogs'), 'accuracy'].max())
# print(realnets_df.loc[(realnets_df['algorithm'] == 'ECG') & (realnets_df['network'] == 'pol_blogs'), 'accuracy'].max())
return realnets_df
##############################################################################
def load_datas():
df = load_df()
dfs_noise = load_dfs_noise(df)
speed = load_speed(df)
df_A = load_df_A(dfs_noise)
realnets_df = load_realnets_df()
return df, dfs_noise, speed, df_A, realnets_df
##############################################################################
## Functions ############################################################################
def create_noises_plot_dir(outname='no_name'):
outdir = './outputs/noises/plots'
if not os.path.exists(outdir):
os.mkdir(outdir)
fullname = os.path.join(outdir, outname)
return fullname
def get_ax(font_size=45, n_cols=4, width=50):
plt.clf()
matplotlib.rc('xtick', labelsize=font_size*.9)
matplotlib.rc('ytick', labelsize=font_size*.9)
return plt.subplots(nrows=1, ncols=n_cols, figsize=(width,10)) # num=None , dpi=250 , figsize=(7.8*4, 6.3)
def save_plot(ax, title='no title', x_label='', y_label='', f_name='',
font_size=45, n_col=1, labels_handles=None, has_legend=True,
font_scale=.8, leg_loc=0, title_loc='center', title_y=-.275):
handles, labels = ax.get_legend_handles_labels()
# handles = []
# for alg in labels:
# handles.append(mpatches.Patch(color=alg_colors[alg], label=alg))
# handles = [mpatches.Patch(color=c, label=key) for key, c in alg_colors.items()]
# labels = list(alg_colors)
if len(labels) == 5:
order = [3,1,2,0,4]
handles = [handles[idx] for idx in order]
labels = [labels[idx] for idx in order]
if labels_handles is not None:
handles = [mpatches.Patch(color=c, label=key) for key, c in labels_handles.items()]
labels = list(labels_handles)
if has_legend:
for h in handles:
h.set_linewidth(5)
leg = ax.legend(fontsize=font_size*font_scale, ncol=n_col, handles=handles, labels=labels, loc=leg_loc)
ax.set_title(title, fontsize=font_size, loc=title_loc, y=title_y)
ax.set_xlabel(x_label, fontsize=font_size)
ax.set_ylabel(y_label, fontsize=font_size)
ax.grid(True)
##############################################################################
## Generate Plots ############################################################################
def generate_noises_plot():
for noise_graph, data in dfs_noise.items():
fig, ax = get_ax(n_cols=1, width=12.5)
ax.set_ylim(.499,1.01)
g = sns.lineplot(x="q/p", y="accuracy", hue="algorithm", data=data, marker='o', linewidth=4.5, ax=ax, ci=95, hue_order=['Louvain','local improve','hedonic','spectral','ECG']) # , style="event"
ax.axhline(1-noise_graph, c='red', alpha=.75, ls='--', lw=2.5)
# ax.annotate('$1$-noise', (.85, 1.005-noise_graph), fontsize=40)
noise_name = str(round(noise_graph,3))+'00000000'
save_plot(ax, title=f'noise={noise_name[:5]}', x_label='q/p', y_label='accuracy', f_name=f'{file_name_prefix}_n{round(noise_graph,3)}', leg_loc='lower left', title_loc='right', title_y=.9)
plt.tight_layout()
fig.savefig(create_noises_plot_dir(f'noise_{noise_graph}.png'))
# break
##############################################################################
def generate_gif():
frames = []
folder = './outputs/noises/plots/'
imgs = [png for png in os.listdir(folder) if png[-3:] == 'png'] # get_all_files()#glob.glob('.outputs/noises/*.png')
imgs = [(img, eval(img[6:-4])) for img in imgs]
imgs.sort(key= lambda x: x[1])
imgs = [img[0] for img in imgs]
for i in imgs:
new_frame = Image.open(folder+i)
frames.append(new_frame)
frames[0].save('noises.gif', format='GIF',
append_images=frames[1:],
save_all=True,
duration=500, loop=0)
##############################################################################
def generate_main_plot():
fig, axes = get_ax()
noise_graph = .35 # Choose one noise to be in Fig 1 (a)
# print(dfs_noise[noise_graph]['algorithm'].unique())
axes[0].set_ylim(.499,1.01)
g = sns.lineplot(x="q/p", y="accuracy", hue="algorithm", data=dfs_noise[noise_graph], marker='o', linewidth=4.5, ci=95, hue_order=['Louvain','local improve','hedonic','spectral','ECG'], ax=axes[0]) # , style="event"
axes[0].axhline(1-noise_graph, c='red', alpha=.75, ls='--', lw=2.5)
axes[0].annotate('$1$-noise', (.85, 1.005-noise_graph), fontsize=40)
plt.text(0.6, 0.65, '', bbox=dict(boxstyle="round", fc="white", ec="red", pad=0.2))
noise_name = str(round(noise_graph,3))+'00000000'
save_plot(axes[0], title=f'(a) accuracy vs q/p (noise={noise_name[:5]})', x_label='q/p', y_label='accuracy', f_name=f'{file_name_prefix}_n{round(noise_graph,3)}')
print('plot 1')
##############################################################################
g = sns.lineplot(x="noise", y="r = maximum tolerated q/p", hue="algorithm", data=df_A, marker='o', linewidth=4.5, hue_order=['Louvain','local improve','hedonic','spectral','ECG'], ax=axes[1])
save_plot(axes[1], title='(b) max q/p such that accuracy$\geq1$-noise', x_label='noise', y_label='r = maximum tolerated q/p', f_name='cross')
print('plot 2')
##############################################################################
# ['Louvain','local improve','hedonic','spectral','ECG']
# order=['spectral','local\nimprove','hedonic','Louvain','ECG']
# ['Louvain' 'ECG' 'hedonic' 'local\nimprove' 'spectral']
# print('ki', speed['algorithm'].unique())
palette=alg_colors.copy()
palette['local\nimprove'] = palette.pop('local improve')
g = sns.violinplot(x="algorithm", y="seconds", data=speed, fontsize=27.5, palette=palette, order=['spectral','local\nimprove','hedonic','Louvain','ECG'], ax=axes[2]) # hue_order=['spectral','local\nimprove','hedonic','Louvain','ECG']
# g = sns.boxplot(x="algorithm", y="seconds", data=speed, ax=axes[2]) # , fontsize=27.5
# g = sns.swarmplot(x="algorithm", y="seconds", data=speed, color=".25", ax=axes[2])
axes[2].set(yscale="log") # xscale="log",
save_plot(axes[2], title='(c) time efficiency', x_label='', y_label='seconds', f_name='time', has_legend=False) # algorithm
print('plot 3')
##############################################################################
alg_repetition = int((len(realnets_df['algorithm'].unique())-3)/2) # #5975A4 sns.color_palette("Blues", n_colors=1)
cor = [(alg_colors['Louvain']),(alg_colors['ECG']),(alg_colors['spectral'])]+sns.color_palette("Oranges", n_colors=alg_repetition)+sns.color_palette("Greens", n_colors=alg_repetition) # #5975A4 #CC8964 #5F9E6E
realnets_order = ['ml','ecg','spectral','local improve_n0','local improve_n0.025','local improve_n0.05','local improve_n0.1','local improve_n0.15','local improve_n0.2','local improve_n0.25','local improve_n0.3','local improve_n0.35','local improve_n0.4','local improve_n0.45','local improve_n0.475','local improve_n0.5','hedonic_n0','hedonic_n0.025','hedonic_n0.05','hedonic_n0.1','hedonic_n0.15','hedonic_n0.2','hedonic_n0.25','hedonic_n0.3','hedonic_n0.35','hedonic_n0.4','hedonic_n0.45','hedonic_n0.475','hedonic_n0.5']
g = sns.barplot(x="network", y="accuracy", hue="algorithm", data=realnets_df, palette=cor, hue_order=realnets_order, ax=axes[3]) # order=realnets_order
max_acc = [] # [0] * len(realnets_order) * 4
for net in ['karate', 'dolphins', 'pol_books', 'pol_blogs']:
nets_df = realnets_df.loc[realnets_df['network'] == net]
for alg in realnets_order: # nets_df['algorithm'].unique():
nets_alg_df = nets_df.loc[nets_df['algorithm'] == alg] # .max()
max_acc.append(nets_alg_df['accuracy'].max())
# max_acc[realnets_order.index(alg)] = nets_alg_df['accuracy'].max()
x_pos = [p.get_x() for p in axes[3].patches]
x_pos.sort()
colors = [alg_colors['Louvain'],alg_colors['ECG'],alg_colors['spectral']]+[alg_colors['local improve'] for _ in range(alg_repetition)]+[alg_colors['hedonic'] for _ in range(alg_repetition)]
for pos, maxx, c in zip(x_pos, max_acc, colors*4):
axes[3].annotate('_', (pos, maxx), c=c, weight=1000) # p.get_height() * 1.005)
axes[3].set_ylim(0.499,1.01)
# labls = {'spectral clustering':'#5975A4', 'local improvement\n(0$\leq$noise$\leq$0.5)':'#CC8964', 'hedonic robust\n(0$\leq$noise$\leq$0.5)':'#5F9E6E'}
labls = {'Louvain':alg_colors['Louvain'], 'ECG':alg_colors['ECG'], 'spectral':alg_colors['spectral'], 'l. improve':alg_colors['local improve'], 'hedonic':alg_colors['hedonic']} # \n(0$\leq$noise$\leq$0.5)
save_plot(axes[3], title='(d) real networks', x_label='network', y_label='accuracy', f_name='real_nets_bar_plot', n_col=2, labels_handles=labls, font_scale=.625)
print('plot 4')
##############################################################################
plt.tight_layout()
fig.savefig('main_plot.png')
print('save fig')
##############################################################################
## Plot Comparison ############################################################################
def filter_gt_balanced(df):
print(df['max_comp'].max(), df['max_comp'].min())
# print(df.columns)
adasdadas()
# df = df[df['gt_balance'] ]
def generate_plot_comparion(filename=''):
df = pd.read_csv(filename)
# df = filter_gt_balanced(df)
plot_hists(df)
asdas()
# print(len(df))
# df['q'] = df['p_in'] * df['mult']
# df = df[(df['p_in']+df['q']) / 2 > np.log2(500*2)] # ( p+q ) / 2 > log n
# print(len(df))
font_size = 45
plt.clf()
matplotlib.rc('xtick', labelsize=font_size*.8)
matplotlib.rc('ytick', labelsize=font_size*.8)
fig, axes = plt.subplots(nrows=2, ncols=4, figsize=(60,30)) # num=None , dpi=250 , figsize=(7.8*4, 6.3)
i, j = 0, 0
methods = df['method'].unique()
for method in methods:
df_method = df[df['method']==method]
ax = axes[i][j]
sns.lineplot(x='mult', y='accuracy', hue='algorithm', data=df_method, marker='o', linewidth=4, ax=ax)
ax.set_title(method, fontsize=font_size)
ax.set_xlabel('q/p', fontsize=font_size*.9)
ax.set_ylabel('accuracy', fontsize=font_size*.9)
ax.grid(True)
handles, labels = ax.get_legend_handles_labels()
# handles, labels = handles[1:], labels[1:]
for h in handles:
h.set_linewidth(5)
ax.legend(fontsize=font_size*.7, handles=handles, labels=labels, loc='upper right')
j += 1
if j == 4:
j = 0
i += 1
ax = axes[i][j]
ax.set_title('speed', fontsize=font_size)
ax.set_xlabel('q/p', fontsize=font_size*.9) # algorithms
ax.set_ylabel('seconds', fontsize=font_size*.9)
# ax. # log scale
# ax.set(ylim=(0, .15))
# sns.violinplot(x='algorithm', y='seconds', data=df, fontsize=27.5, ax=ax) # [df['method']=='rand']
sns.lineplot(x='mult', y='seconds', hue='algorithm', data=df[df['method']=='dist'], marker='o', linewidth=4, ax=ax)
ax.grid(True)
handles, labels = ax.get_legend_handles_labels()
for h in handles:
h.set_linewidth(5)
ax.legend(fontsize=font_size*.7, handles=handles, labels=labels, loc='upper left')
plt.savefig(f'{filename[:-4]}_junto_filter.png')
##########################
## Hedonic Robust vs Naive
def plot_robust_vs_naive():
df = pd.read_csv('outputs/comparisons/rob_vs_naive__ps=10_mults=11_inst=10_reps=10_nComm=2_commSize=50.csv')
plt.clf()
print(np.mean(df[df['algorithm']=='hedonic']['accuracy']))
print(np.mean(df[df['algorithm']=='naive']['accuracy']))
print(len(df))
# sns.jointplot(data=df, x="robustness", y="accuracy", hue="algorithm", kind="scatter") # scatter kde hist hex reg resid
# sns.violinplot(x='algorithm', y='seconds', data=df, fontsize=27.5) # [df['method']=='rand']
sec0 = df[df['algorithm']=='naive']['seconds']
sec0 = [s for s in sec0 if s > .01]
sec1 = df[df['algorithm']=='hedonic']['seconds']
sec1 = [s for s in sec1 if s > .01]
# plt.hist(sec) # [df['method']=='rand']
sns.violinplot(sec0, alpha=.1, color='green') # [df['method']=='rand']
sns.violinplot(sec1, alpha=.1, color='red') # [df['method']=='rand']
plt.show()
##########################
## Python Naive vs Networkx
def plot_python_vs_netx(filename=''):
df = pd.read_csv(filename)
plt.clf()
# matplotlib.rc('xtick', labelsize=font_size*.8)
# matplotlib.rc('ytick', labelsize=font_size*.8)
fig, ax = plt.subplots(dpi=200) # nrows=2, ncols=4, figsize=(60,30) num=None , dpi=250 , figsize=(7.8*4, 6.3)
sns.lineplot(x='mults', y='secs', hue='alg', data=df, marker='o', linewidth=4, ax=ax) # , ax=ax
ax.set_title('time comparison between implementations') # , fontsize=font_size
ax.set_xlabel('q/p') # algorithms , fontsize=font_size*.9
ax.set_ylabel('seconds') # , fontsize=font_size*.9
handles, labels = ax.get_legend_handles_labels()
labels = ['previous','current'] # 'implementations'
for h in handles:
h.set_linewidth(4)
leg = ax.legend(handles=handles, labels=labels) # , loc=leg_loc
plt.savefig(f'{filename[:-4]}_time.png')
################################################################################
def plot_hists(filename):
# ['nodes', 'max_comp', 'edges', 'gt_balance', 'p_in', 'mult', 'instance', 'repetition', 'algorithm', 'accuracy', 'robustness', 'method', 'seconds']
df = pd.read_csv(filename)
plt.clf()
plt.hist(df['gt_balance'])
plt.savefig('gt_balance.png')
plt.clf()
plt.hist(1-(df['max_comp']/df['nodes']))
plt.savefig('isolated_nodes.png')
plt.clf()
plt.hist(df['edges'] / (df['max_comp']*(df['max_comp']-1)/2))
plt.savefig('edge_density.png')
##############################################################################
## Main ############################################################################
if __name__ == "__main__":
# df = load_df()
# df = load_realnets_df()
# realnets_df = load_realnets_df()
# print(realnets_df['algorithm'].unique())
# asdasd()
# print(realnets_df.columns)
# nets = realnets_df['network'].unique()
# for net in nets:
# print(net, np.mean(realnets_df.loc[(realnets_df['network'] == net) & (realnets_df['algorithm'] == 'ecg')]['accuracy']))
# df = load_df()
# speed = load_speed(df)
# accs_ecg = speed[speed['algorithm'] == 'ECG']['seconds'].values()
# print(np.mean(speed[speed['algorithm'] == 'ECG']['seconds']))
# for mult in speed['mult'].unique():
# plt.clf()
# plt.title(mult)
# plt.hist(speed.loc[(speed['mult'] == mult) & (speed['algorithm'] == 'ECG')]['seconds'])
# plt.show()
df, dfs_noise, speed, df_A, realnets_df = load_datas()
# generate_noises_plot()
# generate_gif()
generate_main_plot()
# name = 'max_components__ps=10_mults=11_inst=10_reps=10_nComm=2_commSize=500.csv'
# fname = f'outputs/comparisons/news/{name}'
# generate_plot_comparion(fname)
# plot_hists(fname)
# df = pd.read_csv('outputs/comparisons/comparison_commSize=111.csv')
# for alg in df['algorithm'].unique():
# plt.clf()
# plt.violinplot(df[df['algorithm']==alg]['seconds'])
# plt.savefig(f'{alg}.png')
# sns.violinplot(x='algorithm', y='seconds', data=df[df['method']=='rand'], fontsize=27.5, ax=ax)
# plot_robust_vs_naive()
# plot_python_vs_netx('outputs/speed_python_vs_netx/speed_test__ps=5_mults=6_inst=10_reps=10_nComm=2_commSize=250.csv')