-
Notifications
You must be signed in to change notification settings - Fork 0
/
Module_missing_labels.py
284 lines (196 loc) · 11 KB
/
Module_missing_labels.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
import copy
from EBLS import *
import numpy as np
import time
from sklearn.preprocessing import normalize
from numpy.random import default_rng
class Module_missing_labels():
def __init__(self,label_count,N1,N2,N3,max_learners, data_numbers):
self.data_numbers = data_numbers
self.max_learners = max_learners
self.N1 = N1
self.N2 = N2
self.N3 = N3
self.s = 0.8 #shrink coefficient
self.C = 2**-30 # Regularization coefficient
self.learners = []
self.acc_BLS = EBLS()
self.acc_BLS_2 = EBLS()
self.acc_BLS_3 = EBLS()
self.label_count = label_count
self.ensemble_accs_labeled = []
self.initialized = False
self.beta_1 = []
self.learners_count = 0
self.preds_w = np.zeros((self.data_numbers,self.label_count))
self.ensemble_accs = []
self.current_learner = 0
self.worst_list = []
self.restarted_list = []
self.eleminated = []
self.old_besties = []
self.threshold = 50
def test_then_train(self,test_whole,is_binary,x_list, y_list,x_list_labeled,y_list_labeled,y_list_missing,missing_data,missing_label,missing_label_all):
test_whole = np.array(test_whole)
removed_instances_all = np.where(missing_label_all == 2)[0]
removed_instances = np.where(y_list_missing == 2)[0]
removed_instances_all= np.unique(removed_instances_all)
new_TT3 = test_whole.copy()
if(is_binary):
if(len(new_TT3) > 0):
new_TT3 = np.delete(new_TT3, removed_instances, axis=0)
yy = y_list_labeled.copy()
x_list_labeled = np.delete(x_list_labeled, removed_instances, axis=0 )
yy = np.delete(yy, removed_instances, axis=0)
y_list_labeled =yy.copy()
else:
new_TT3 = test_whole.copy()
y_list_labeled= missing_label_all
p_labeled = []
ensemble_acc_2 = -1
ensemble_acc = -1
tmp_batch_acc = -1
preds = np.zeros((x_list.shape[0],self.label_count))
pred_prob = np.zeros((x_list.shape[0],self.label_count))
pred_prob_w = np.zeros((len(x_list),self.label_count))
if(self.learners_count < self.max_learners ):
EBLS_tmp = None
EBLS_tmp = EBLS()
if(is_binary == False):
print("here")
self.restarted_list = []
self.learners.append(EBLS_tmp)
self.current_learner = self.learners_count# +
self.learners_count += 1
self.restarted_list.append(self.current_learner)
if(self.learners_count >= self.max_learners or (len(self.worst_list) > len(self.learners) /2)):
self.restarted_list = []
if(len(self.worst_list) != 0):
for k in range(0 ,int(((len(self.worst_list))) -1 )):
l = self.worst_list[k]
if(len(self.worst_list) > len(self.learners)/2 and l!=0):
self.eleminated.append(self.learners[l])
if(l!=0 ):
self.learners[l] = 0
learners_tmp_new = []
for k in range(0 , len(self.learners)):
if(self.learners[k] != 0):
learners_tmp_new.append(self.learners[k])
self.learners = learners_tmp_new
ff = 0
while (len(self.learners) < (self.max_learners ) and ff < len(self.old_besties)):
if(self.old_besties[ff] < len(self.eleminated)):
tmpppp = copy.deepcopy(self.eleminated[self.old_besties[ff]])
self.learners.append(tmpppp)
del self.eleminated[self.old_besties[ff]]
self.restarted_list = []
ff += 1
if(len(self.eleminated ) > 100):
self.eleminated = self.eleminated[20:100]
self.old_besties = []
self.learners_count = len(self.learners)
self.worst_list = []
if(self.initialized == False ):
self.beta_1 = x_list_labeled
for m in range(0 , self.learners_count):
self.learners[m].update_second_layer(self.beta_1,y_list_labeled,self.C)
self.initialized = True
elif(self.initialized == True ):
TT3_w = test_whole#self.BLS.test_first_phase(self.x_list)
self.preds_w = np.zeros((len(x_list),self.label_count))
aa = time.time()
for m in range(0 , self.learners_count):
if((m not in self.restarted_list) and len(self.learners[m].beta2 ) > 0):
if(len(y_list_labeled)!= 0):
self.my_accs ,batc_acc ,m_pred = self.learners[m].test_second_phase(new_TT3,y_list_labeled)
else:
batc_acc =100
m_pred_w = []
#m_pred_w = m_pred
m_pred = self.learners[m].just_test_second_phase(TT3_w)
m_pred_w = m_pred
if(len(self.ensemble_accs_labeled)!=0):
self.threshold = self.ensemble_accs_labeled[len(self.ensemble_accs_labeled)-1]
if(batc_acc < self.threshold) :
self.worst_list.append(m)
prediction_prob = m_pred.copy()
prediction = m_pred.copy()
prediction_prob_w = m_pred_w.copy()
prediction_w = m_pred_w.copy()
prediction_mins = np.zeros_like(prediction)
prediction_mins_w = np.zeros_like(prediction_w)
prediction_mins[np.arange(len(prediction)), prediction.argmax(1)] = 1
prediction_mins_w[np.arange(len(prediction_w)), prediction_w.argmax(1)] = 1
prediction = 1 * prediction_mins
prediction_w = 1 * prediction_mins_w
preds = preds + prediction
pred_prob = pred_prob +prediction_prob
self.preds_w = self.preds_w + prediction_w
pred_prob_w = pred_prob_w +prediction_prob_w
for jj in range(0 , len(preds)):
if(preds[jj,0] == preds[jj,1]):
preds[jj] = pred_prob[jj]
for jj in range(0 , len(self.preds_w)):
if(self.preds_w[jj,0] == self.preds_w[jj,1]):
self.preds_w[jj] = pred_prob_w[jj]
self.old_besties = []
for ll in range(0 , len(self.eleminated)):
if(len(y_list_labeled) != 0):
my_accs_e ,batc_acc_e ,m_pred_e = self.eleminated[ll].test_second_phase(new_TT3,y_list_labeled)
else:
batc_acc_e = 0
if(is_binary == True):
th = 50
else:
th = 30
if(batc_acc_e > th):
self.old_besties.append(ll)
indexes_2 = []
indexes = []
if(is_binary):
indexes, ensemble_acc , tmp_batch_acc , wrong_index = self.acc_BLS.show_accuracy_ensemble(self.preds_w,y_list,len(x_list))
preds_copy = self.preds_w.copy()
preds_copy = np.delete(preds_copy, removed_instances, axis=0)
if(len(y_list_labeled)!= 0):
indexes_2, ensemble_acc_2 , tmp_batch_acc_2 , wrong_index_2 = self.acc_BLS_2.show_accuracy_ensemble(preds_copy,y_list_labeled,len(x_list_labeled))
else:
ensemble_acc_2= 100
if( is_binary == False ): #non_binary
self.preds_w = self.preds_w / (len(self.learners)-1)
self.preds_w = normalize(self.preds_w, axis=1, norm='max')
for k in range(0, len(self.preds_w)):
iindex = np.argmax(self.preds_w[k])
self.preds_w[k,iindex] = 1
for l in range(0, self.label_count):
if(l != iindex):
self.preds_w[k,l] = 0
_, ensemble_acc , tmp_batch_acc , wrong_index = self.acc_BLS_3.show_accuracy_ensemble_multi_label(self.preds_w,y_list,len(x_list))
ensemble_acc_2 = ensemble_acc
ensemble_acc = ensemble_acc * 100
ensemble_acc_2 = ensemble_acc_2 * 100
self.ensemble_accs.append(ensemble_acc)
self.ensemble_accs_labeled.append(ensemble_acc_2)
p_labeled = pred_prob_w.copy()
if(is_binary and len(indexes_2) > 0):
#print("oops")
#time.sleep(2)
y_tmp = y_list_labeled[indexes_2]
x_tmp = x_list_labeled[indexes_2]
x_tmp = np.array(x_tmp)
y_tmp = np.array(y_tmp)
x_list_labeled = np.vstack((x_list_labeled,x_tmp))
y_list_labeled = np.vstack((y_list_labeled,y_tmp))
self.beta_1 = x_list_labeled
learners_choice = []
rng = default_rng()
for zz in range(0 , len(self.learners)):
second_phase_start = time.time()
if(len(y_list_labeled)!=0):
self.learners[zz].update_second_layer(self.beta_1 ,y_list_labeled,self.C)
self.preds_w = np.zeros((self.data_numbers,self.label_count))
pred_prob_w = np.zeros((self.data_numbers,self.label_count))
preds = np.zeros((y_list.shape[0],self.label_count))
pred_prob = np.zeros((y_list.shape[0],self.label_count))
self.x_list = []
y_list = []
return p_labeled,tmp_batch_acc