-
Notifications
You must be signed in to change notification settings - Fork 1
/
test_real_data.jl
676 lines (587 loc) · 21.7 KB
/
test_real_data.jl
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
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
#-------------------------------------------------------------------------------
# Test the code on real data
# In the 'Data' section, chose the name of the dataset of interest in
# "data_name". The file must then be in data/$data_name.txt.
# The results will then be stored in results/$data_name/$results_path/time_of_run
# Set the parameters of the Gibbs sampler in the 'Set Parameters'
# section.
#-------------------------------------------------------------------------------
include("src/main.jl")
using ProgressMeter
using HDF5, JLD
#-------------------------------------------------------------------------------
# Data
#-------------------------------------------------------------------------------
#data_name = "Email-Enron"
#data_name = "polblogs"
#data_name = "wikipedia_3000"
#data_name = "NIPS234"
#data_name = "deezer_RO_und"
data_name ="data_for_sigma_800"
# Result folder name
result_folder = "sigma 025"
# Comments about the run
comments = "Alpha free\n new update of mu tilde \n new measure unobs \n sigma max 0.25"
#-------------------------------------------------------------------------------
# Set parameters
#-------------------------------------------------------------------------------
kappa=1.;tau=10.;sigma=0.;alpha=.04;beta=.5
warm_start = false
weighted = false
directed = true
# Are there self edges
self_edge = true
# Proportion of entries to mask
pred_ratio = 0.
# Saving clustering
save_clustering = true
# Number of clustering to save
n_clusterings = 50
n_clusterings = (save_clustering ? n_clusterings : 2)
# Set to true to display the current value of sigma every 5% of the progress. It also
# allows to save the current state of the run every 5%. Then use continue_real.jl if the script
# is stopped before the end of the sampler to carry from last save.
monitoring_sigma = true
# Number of iterations for the Gibbs sampler
n_iter = 10000
# Thinning the chain, store only every skip iteration
skip = 50
# Number of iterations as burn-in
burn = floor(Int,2.*n_iter/4)
# Number of activities to store
K = 10
# Number of updates of the hyperparameters per iteration
n_steps_hyper = 10
# Hyperparameters for the MH update
prior_params = Dict()
prior_params["kappa"] = (.1,.1)
prior_params["sigma"] = (.1,.1)
prior_params["tau"] = (.1,.1)
prior_params["alpha"] = (.01,.1)
prior_params["beta"] = (.1,.1)
prop_params = Dict()
prop_params["kappa"] = 0.04
prop_params["sigma"] = 0.08
prop_params["tau"] = 0.04
prop_params["alpha"] = 0.04
prop_params["beta"] = 0.04
# Set to true if the parameter is fixed by user (to the initial value)
FIXED_KAPPA = false
FIXED_SIGMA = false
FIXED_TAU = false
FIXED_ALPHA = false
FIXED_BETA = true
#-------------------------------------------------------------------------------
# Importing data
#-------------------------------------------------------------------------------
file_name = string("data/",data_name,".txt")
println(string("Reading the file ",file_name))
@time A = readdlm(file_name,Float64)
println()
println(comments)
println()
# Convert in sparse matrix format
println("Creating sparse matrix")
i_data = [trunc(Int,x) for x in A[:,1]]
j_data = [trunc(Int,x) for x in A[:,2]]
if weighted
val_data = A[:,3]
else
val_data = ones(Int, length(i_data))
end
n = max(maximum(i_data),maximum(j_data))
sparse_data = sparse(i_data,j_data,val_data,n,n)
# If the data is undirected, keep only below diagonal
if directed==false
for t in 1:length(i_data)
i_ = i_data[t]
j_ = j_data[t]
if i_ < j_
if weighted
sparse_data[j_,i_] += sparse_data[i_,j_]
sparse_data[i_,j_] = 0
else
sparse_data[j_,i_] = 1
sparse_data[i_,j_] = 0
end
end
end
end
# Make sure there are no self edges if self_edge is set to true
if self_edge == false
for t in 1:length(i_data)
i_ = i_data[t]
j_ = j_data[t]
if i_ == j_
sparse_data[i_,j_] = 0
end
end
end
sparse_data = dropzeros(sparse_data)
println()
#-------------------------------------------------------------------------------
# Save folder
#-------------------------------------------------------------------------------
# Path to the folder where to store the information
main_dir = pwd()
current_dir = Dates.format(now(),"dd-mm-yy_HH-MM-SS")
# Saving main informations about the run in a txt file
println("Saving in "*current_dir)
println()
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,'/')
mkpath(results_path)
open(results_path*"info.txt","w") do f
write(f,"Dataset:\n")
write(f,string(" name = ", data_name,"\n"))
write(f,string(" number of nodes = ", n,"\n"))
write(f,string(" number of edges = ", sum(sparse_data),"\n"))
write(f,string(" directed = ", directed,"\n"))
write(f,string(" weighted = ", weighted,"\n\n"))
write(f,"Initial parameters:\n")
write(f,string(" kappa = ", kappa,"\n"))
write(f,string(" tau = ", tau,"\n"))
write(f,string(" sigma = ", sigma,"\n"))
write(f,string(" alpha = ", alpha,"\n"))
write(f,string(" beta = ", beta,"\n"))
write(f,string(" warm_start = ", warm_start,"\n\n"))
write(f,"Hyper parameters:\n")
write(f,string(" kappa = ", prior_params["kappa"],"\n"))
write(f,string(" tau = ", prior_params["tau"],"\n"))
write(f,string(" sigma = ", prior_params["sigma"],"\n"))
write(f,string(" alpha = ", prior_params["alpha"],"\n"))
write(f,string(" beta = ", prior_params["beta"],"\n\n"))
write(f,string("Number of iterations = ", n_iter*skip,"\n"))
write(f,string("Comments : ", comments,"\n"))
end
#-------------------------------------------------------------------------------
# Initializing parameters
#-------------------------------------------------------------------------------
n=first(size(sparse_data))
# List of the number of active communities
n_active_list = zeros(Int,n_iter)
# List of the top K activites
activities_list = zeros(n_iter,K)
# Lists of values of the parameters at each step
kappa_list = zeros(n_iter)
sigma_list = zeros(n_iter)
tau_list = zeros(n_iter)
alpha_list = zeros(n_iter)
beta_list = zeros(n_iter)
# Monitoring s_min through the run
s_min_list = zeros(n_iter)
# List of l2 errors for prediction
error_list = zeros(n_iter)
error_mean_list = zeros(n_iter)
error_mean_list_burn = zeros(n_iter)
s_min = 0.
# Initializing variables
R_ = Array{Float64,1}()
V_ = Affinity()
# Initialize clusterings variables
# Initialize index of current clustering
idx_clustering = 1
# Iterations for which we save the clustering
ind_clusterings = zeros(Int,n_clusterings)
# Save the assignation of each node to its cluster
# clusterings[i,clus_idx] = idx of cluster of node i
clusterings = zeros(Int,n,n_clusterings)
# Initialize current values of parameters
c_kappa = kappa
c_sigma = sigma
c_tau = tau
c_alpha = alpha
c_beta = beta
# Debugging variables
plot_true = false
PRINT_ = false
#-------------------------------------------------------------------------------
# Initializing variables for predictions
#-------------------------------------------------------------------------------
# Observed matrix
Z_tilde = copy(sparse_data)
# Select entries to predict
println("Masking entries to predict")
if directed
n_to_predict = trunc(Int,pred_ratio*(n^2-n))
else
n_to_predict = trunc(Int,pred_ratio/2*(n^2-n))
end
println("Select $n_to_predict indices to predict")
# First select more than needed entries, since we can't hide
# any entry
if directed
n_to_predict = trunc(Int,1.15*pred_ratio*n^2)
else
n_to_predict = trunc(Int,1.15*pred_ratio/2*n^2)
end
I_pred,J_pred = rand(1:n,n_to_predict),rand(1:n,n_to_predict)
if directed == false
for t in 1:n_to_predict
i_pred = I_pred[t]
j_pred = J_pred[t]
if i_pred > j_pred
I_pred[t] = j_pred
J_pred[t] = i_pred
end
end
end
@time to_predict = sparse(I_pred,J_pred,ones(Int64,n_to_predict),n,n)
I_pred,J_pred = findnz(to_predict)
n_to_predict = length(I_pred)
v_true = Array{Int64}(n_to_predict)
is_test = zeros(Int64,n_to_predict)
println("Mask corresponding entries")
@time for t in 1:n_to_predict
i_pred = I_pred[t]
j_pred = J_pred[t]
if i_pred != j_pred
if Z_tilde[i_pred,j_pred] == 0 || ( sum(Z_tilde[i_pred,:]) > 1 && sum(Z_tilde[:,j_pred]) > 1 )
is_test[t] = 1
v_true[t] = sparse_data[i_pred,j_pred]
if weighted == false
v_true[t] = min(v_true[t],1)
end
Z_tilde[i_pred,j_pred] = 0
end
end
if sum(is_test) > pred_ratio/2*(n^2-n) && directed == false
break
end
if sum(is_test) > pred_ratio*(n^2-n)
break
end
end
n_to_predict = sum(is_test)
to_predict = sparse(I_pred[find(is_test)], J_pred[find(is_test)], ones(Int64,n_to_predict),n,n)
I_pred,J_pred = findnz(to_predict)
v_true = v_true[find(is_test)]
I_tilde,J_tilde,V_tilde = findnz(Z_tilde)
Z_tilde = sparse(I_tilde,J_tilde,V_tilde,n,n)
# Vecor with integer predictions
pred_vect = ones(Float64,n_to_predict)
# Vector of posterior mean
pred_average_vect = zeros(Float64,n_to_predict)
pred_average_vect_burn = zeros(Float64,n_to_predict)
# Sparse matrix of observed entries and the ones to predict
I_all,J_all,V_all = findnz(Z_tilde+to_predict+speye(Int64,n))
all_ind_mat = dropzeros(sparse(I_all,J_all,ones(Int,length(I_all)),n,n))
println()
if warm_start == false
partition_ = Factorized{Bool}()
sentAndReceived_ = Count()
partition_[1] = Z_tilde
sentAndReceived_[1] = reshape(sum(Z_tilde,1),n) + reshape(sum(Z_tilde,2),n)
else
K_init = trunc(Int,active_feature_mean(n, c_kappa, c_tau, c_sigma, c_alpha, c_beta))+1
s_min_init = Inf
r_dist_init = Gamma(1.-c_sigma,1./c_tau)
R_ = zeros(K_init)
for k in 1:K_init
R_[k] = rand(r_dist_init)
V_[k] = c_alpha/c_beta*ones(n)
s_min_init = min(s_min_init,R_[k])
end
slice_matrix = s_min_init*all_ind_mat
if weighted
partition_,sentAndReceived_ = update_partition(R_,V_,slice_matrix,Z_tilde,to_predict,pred_vect,directed,self_edge)
else
partition_,sentAndReceived_ = update_partition_unweighted(R_,V_,slice_matrix,Z_tilde,to_predict,pred_vect,directed,self_edge)
end
end
#-------------------------------------------------------------------------------
# Starting the MCMC iterations
#-------------------------------------------------------------------------------
println(string("Starting Gibbs sample with ",n_iter," steps"))
tic()
@profile @showprogress for i in 1:n_iter
#FIXED_SIGMA = (i < n_iter/4)
# Update measure
R_,V_,n_observed,slice_matrix,s_min = update_measure(partition_,sentAndReceived_,all_ind_mat,c_kappa,c_tau,c_sigma,c_alpha,c_beta)
# Update partition
if weighted
partition_,sentAndReceived_ = update_partition(R_,V_,slice_matrix,Z_tilde,to_predict,pred_vect,directed,self_edge)
else
partition_,sentAndReceived_ = update_partition_unweighted(R_,V_,slice_matrix,Z_tilde,to_predict,pred_vect,directed,self_edge)
end
# Update hyperparameters
for t in 1:n_steps_hyper
c_kappa,c_sigma,c_tau,c_alpha,c_beta = update_parameters_neg2(c_kappa,
c_sigma,
c_tau,
c_alpha,
c_beta,
prior_params,
prop_params,
R_,
V_,
sentAndReceived_,
s_min)
end
PRINT_ = false
for bob in 1:(skip-1)
# Update measure
R_,V_,n_observed,slice_matrix,s_min = update_measure(partition_,sentAndReceived_,all_ind_mat,c_kappa,c_tau,c_sigma,c_alpha,c_beta)
# Update partition
if weighted
partition_,sentAndReceived_ = update_partition(R_,V_,slice_matrix,Z_tilde,to_predict,pred_vect,directed,self_edge)
else
partition_,sentAndReceived_ = update_partition_unweighted(R_,V_,slice_matrix,Z_tilde,to_predict,pred_vect,directed,self_edge)
end
# Update hyperparameters
for t in 1:n_steps_hyper
c_kappa,c_sigma,c_tau,c_alpha,c_beta = update_parameters_neg2(c_kappa,
c_sigma,
c_tau,
c_alpha,
c_beta,
prior_params,
prop_params,
R_,
V_,
sentAndReceived_,
s_min)
end
end
# Store values
n_active_list[i] = n_observed
sort_idx = sortperm(R_,rev=true)
sorted_R_ = R_[sort_idx]
for j in 1:min(n_observed,K)
activities_list[i,j] = sorted_R_[j]*sum(V_[sort_idx[j]])
end
# Store clustering
if i > burn && (i-burn)%floor(Int,(n_iter-burn)/n_clusterings) == 0 && idx_clustering <= n_clusterings
ind_clusterings[idx_clustering] = i
for node_idx in 1:n
weights = [sqrt(R_[c])*V_[c][node_idx] for c in sort_idx[1:min((2*K),length(R_))]]
clusterings[node_idx,idx_clustering] = indmax(weights)
end
idx_clustering += 1
end
pred_average_vect = (i*pred_average_vect+pred_vect)/(i+1)
# Compute auc error (l2 error in comments)
error_list[i] = auc_pr(pred_vect,v_true)#norm(v_true-pred_vect)
error_mean_list[i] = auc_pr(pred_average_vect,v_true)#norm(v_true-pred_average_vect)
i_b = i-burn
if i_b >= 0
pred_average_vect_burn = (i_b*pred_average_vect_burn+pred_vect)/(i_b+1)
error_mean_list_burn[i] = auc_pr(pred_average_vect_burn,v_true)#norm(v_true-pred_average_vect)
end
print_each = floor(Int,n_iter/20)
# If monitoring sigma
if monitoring_sigma && i%print_each == 0
println(string("Progress: ", 100.*i/n_iter,"%"))
println(string("Current sigma: ",c_sigma))
println(string("Length R: ",length(R_)))
# Save Variables
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,"/variables/")
mkpath(results_path)
cd(results_path)
# Store the variables
save("variables.jld","activities_list",activities_list,
"n_active_list",n_active_list,
"kappa_list",kappa_list,
"sigma_list",sigma_list,
"tau_list",tau_list,
"alpha_list",alpha_list,
"beta_list",beta_list,
"partition_",partition_,
"sentAndReceived_",sentAndReceived_,
"clusterings",clusterings,
"ind_clusterings",ind_clusterings,
"n_iter",n_iter,
"skip",skip,
"prop_params",prop_params,
"prior_params",prior_params)
cd(main_dir)
#PRINT_ = true
end
kappa_list[i] = c_kappa
sigma_list[i] = c_sigma
tau_list[i] = c_tau
alpha_list[i] = c_alpha
beta_list[i] = c_beta
end
# Compute partition and corresponding active communities of last iteration
R_t,V_t,n_observed,slice_matrix,s_min = update_measure(partition_,sentAndReceived_,all_ind_mat,c_kappa,c_tau,c_sigma,c_alpha,c_beta)
R_ = R_t[1:n_observed]
V_ = Affinity()
for k in 1:n_observed
V_[k] = V_t[k]
end
elapsed_time = toc()
#-------------------------------------------------------------------------------
# Saving main variables
#-------------------------------------------------------------------------------
# Saving main informations about the run
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,'/')
mkpath(results_path)
open(results_path*"info.txt","w") do f
write(f,"Dataset:\n")
write(f,string(" name = ", data_name,"\n"))
write(f,string(" number of nodes = ", n,"\n"))
write(f,string(" number of edges = ", sum(sparse_data),"\n"))
write(f,string(" directed = ", directed,"\n"))
write(f,string(" weighted = ", weighted,"\n\n"))
write(f,"Initial parameters:\n")
write(f,string(" kappa = ", kappa,"\n"))
write(f,string(" tau = ", tau,"\n"))
write(f,string(" sigma = ", sigma,"\n"))
write(f,string(" alpha = ", alpha,"\n"))
write(f,string(" beta = ", beta,"\n"))
write(f,string(" warm_start = ", warm_start,"\n\n"))
write(f,string("Number of iterations = ", n_iter*skip,"\n"))
write(f,string("Time = ", elapsed_time," s \n\n"))
write(f,string("Comments : ", comments,"\n"))
end
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,"/variables/")
mkpath(results_path)
cd(results_path)
# Store the variables
save("variables.jld","activities_list",activities_list,
"n_active_list",n_active_list,
"kappa_list",kappa_list,
"sigma_list",sigma_list,
"tau_list",tau_list,
"alpha_list",alpha_list,
"beta_list",beta_list,
"partition_",partition_,
"sentAndReceived_",sentAndReceived_,
"clusterings",clusterings,
"ind_clusterings",ind_clusterings,
"n_iter",n_iter,
"skip",skip,
"prop_params",prop_params,
"prior_params",prior_params)
cd(main_dir)
# Load the data using
# load("prediction.jld")[name of the variable]
#-------------------------------------------------------------------------------
# Save prediction
#-------------------------------------------------------------------------------
while true
println()
println("Save prediction vector ? [y/n]")
continue_ = chomp(readline())
if continue_ == "n"
break
end
if continue_ == "y"
plot_true = false
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,"/prediction/")
mkpath(results_path)
cd(results_path)
save("prediction.jld","v_true",v_true,
"predicted",pred_average_vect,
"predicted_burn",pred_average_vect_burn)
# Load the data using
# load("prediction.jld")["v_true"] for true values
# load("prediction.jld")["predicted"] for prediction without burn in
# load("prediction.jld")["predicted_burn"] for prediction with burn in
break
end
end
cd(main_dir)
#-------------------------------------------------------------------------------
# Plotting results
#-------------------------------------------------------------------------------
while true
println()
println("Plot and save results ? [y/n]")
continue_ = chomp(readline())
if continue_ == "n"
break
end
if continue_ == "y"
plot_true = false
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,"/img/")
mkpath(results_path)
cd(results_path)
include(main_dir*"/src/plot_results.jl")
break
end
end
cd(main_dir)
#-------------------------------------------------------------------------------
# Plotting clusters
#-------------------------------------------------------------------------------
while true
println()
println("Plot clusters ? [y/n]")
continue_ = chomp(readline())
if continue_ == "n"
break
end
if continue_ == "y"
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,"/img/")
mkpath(results_path)
cd(results_path)
order,clusters = cluster_communities(R_,V_)
clusters_o = zeros(n)
for (k,cl) in clusters
clusters_o[cl] = k
end
ioff()
spy_sparse_order(sparse_data,order,2.,directed)
spy_sparse_den(sparse_data,clusters_o)
PyPlot.close()
ion()
break
end
end
cd(main_dir)
#-------------------------------------------------------------------------------
# Saving clusters
#-------------------------------------------------------------------------------
while true
println()
println("Save clusters in .jld ? [y/n]")
continue_ = chomp(readline())
if continue_ == "n"
break
end
if continue_ == "y"
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,"/clusters/")
mkpath(results_path)
cd(results_path)
include("src/save_clusters.jl")
break
end
end
cd(main_dir)
#-------------------------------------------------------------------------------
# Saving edge allocation
#-------------------------------------------------------------------------------
order,clusters = cluster_communities(R_,V_)
while true
println()
println("Save edge allocation ? [y/n]")
continue_ = chomp(readline())
if continue_ == "n"
break
end
if continue_ == "y"
# Save partition with posterior order
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,"/img/edge_partition_/")
mkpath(results_path)
cd(results_path)
ioff()
for k in keys(partition_)
spy_sparse_order(partition_[k],order,2.,directed,"Feature $k.png")
PyPlot.close()
end
cd(main_dir)
# Save partition with natural order
results_path = string("results/",data_name,"/",result_folder,'/',current_dir,"/img/edge_partition/")
mkpath(results_path)
cd(results_path)
for k in keys(partition_)
spy_sparse_order(partition_[k],1:n,2.,directed,"Feature $k.png")
PyPlot.close()
end
ion()
break
end
end
PyPlot.close("all")
cd(main_dir)